Power consumption raised significantly in natty

[James Ferguson]

This is a regression moving from maverick->natty. With the same sort of light interactive usage, Battery life has gone from ~4 hours to 2:45 or so.

Running powertop shows the system in P-state "Turbo mode" 10-15% of the time. Typically this was almost zero in maverick, 98+% of the time maverick was in the lowest state.

wakeups reported in the 5-600/s range where previously (from memory) it was 200 ish. i915 driver is high on the list. This is irrespective of whether running unity or classic desktop.

ProblemType: Bug
DistroRelease: Ubuntu 11.04
Package: linux-image-2.6.38-8-generic 2.6.38-8.42
Regression: Yes
Reproducible: Yes
ProcVersionSignature: Ubuntu 2.6.38-8.42-generic 2.6.38.2
Uname: Linux 2.6.38-8-generic x86_64
AlsaVersion: Advanced Linux Sound Architecture Driver Version 1.0.23.
Architecture: amd64
ArecordDevices:
 **** List of CAPTURE Hardware Devices ****
 card 0: Intel [HDA Intel], device 0: CONEXANT Analog [CONEXANT Analog]
   Subdevices: 1/1
   Subdevice #0: subdevice #0
AudioDevicesInUse:
 USER PID ACCESS COMMAND
 /dev/snd/controlC0: james 2068 F.... pulseaudio
CRDA: Error: [Errno 2] No such file or directory
Card0.Amixer.info:
 Card hw:0 'Intel'/'HDA Intel at 0xf2620000 irq 47'
   Mixer name : 'Conexant CX20561 (Hermosa)'
   Components : 'HDA:14f15051,17aa20ff,00100000'
   Controls : 16
   Simple ctrls : 8
Card29.Amixer.info:
 Card hw:29 'ThinkPadEC'/'ThinkPad Console Audio Control at EC reg 0x30, fw 7XHT24WW-1.06'
   Mixer name : 'ThinkPad EC 7XHT24WW-1.06'
   Components : ''
   Controls : 1
   Simple ctrls : 1
Card29.Amixer.values:
 Simple mixer control 'Console',0
   Capabilities: pswitch pswitch-joined penum
   Playback channels: Mono
   Mono: Playback [on]
Date: Wed Apr 13 15:03:35 2011
EcryptfsInUse: Yes
HibernationDevice: RESUME=UUID=bb479652-f524-4abe-b1b0-27646d6deebc
InstallationMedia: Ubuntu 10.10 "Maverick Meerkat" - Release Candidate amd64 (20100928)
MachineType: LENOVO 7465CTO
ProcEnviron:
 LANGUAGE=en_US:en_GB:en
 PATH=(custom, user)
 LANG=en_US.UTF-8
 SHELL=/bin/bash
ProcKernelCmdLine: BOOT_IMAGE=/boot/vmlinuz-2.6.38-8-generic root=UUID=266abe9a-19c2-4cc3-9ef7-238b729b6044 ro quiet splash libata.force=noncq vt.handoff=7
RelatedPackageVersions:
 linux-restricted-modules-2.6.38-8-generic N/A
 linux-backports-modules-2.6.38-8-generic N/A
 linux-firmware 1.50
RfKill:
 0: phy0: Wireless LAN
  Soft blocked: no
  Hard blocked: no
SourcePackage: linux
UpgradeStatus: Upgraded to natty on 2011-04-06 (7 days ago)
dmi.bios.date: 11/10/2009
dmi.bios.vendor: LENOVO
dmi.bios.version: 6DET61WW (3.11 )
dmi.board.name: 7465CTO
dmi.board.vendor: LENOVO
dmi.board.version: Not Available
dmi.chassis.asset.tag: No Asset Information
dmi.chassis.type: 10
dmi.chassis.vendor: LENOVO
dmi.chassis.version: Not Available
dmi.modalias: dmi:bvnLENOVO:bvr6DET61WW(3.11):bd11/10/2009:svnLENOVO:pn7465CTO:pvrThinkPadX200s:rvnLENOVO:rn7465CTO:rvrNotAvailable:cvnLENOVO:ct10:cvrNotAvailable:
dmi.product.name: 7465CTO
dmi.product.version: ThinkPad X200s
dmi.sys.vendor: LENOVO

=== Release Notes ===

Due to a regression inherited from the upstream linux kernel, Natty 11.04 can exhibit a 10-30% increase in power consumption. This is a known issue which is actively being investigated. A SRU will be released onced a viable solution is found. For more information, please see
http://www.phoronix.com/scan.php?page=article&item=linux_mobile_uffda&num=1

[nanog (sorenimpey)]

I am seeing the same behaviour after my upgrade to natty. Battery life has been cut by 40%. Powertop shows 500-700 wake ups per second while in maverick there were typically 200-400 wake ups per second.

2.6.38-8-generic #42-Ubuntu SMP Mon Apr 11 03:31:50 UTC 2011 i686 i686 i386 GNU/Linux

00:00.0 Host bridge: Intel Corporation Mobile 945GME Express Memory Controller Hub (rev 03)
00:02.0 VGA compatible controller: Intel Corporation Mobile 945GME Express Integrated Graphics Controller (rev 03)
00:02.1 Display controller: Intel Corporation Mobile 945GM/GMS/GME, 943/940GML Express Integrated Graphics Controller (rev 03)
00:1b.0 Audio device: Intel Corporation N10/ICH 7 Family High Definition Audio Controller (rev 02)
00:1c.0 PCI bridge: Intel Corporation N10/ICH 7 Family PCI Express Port 1 (rev 02)
00:1c.1 PCI bridge: Intel Corporation N10/ICH 7 Family PCI Express Port 2 (rev 02)
00:1c.2 PCI bridge: Intel Corporation N10/ICH 7 Family PCI Express Port 3 (rev 02)
00:1d.0 USB Controller: Intel Corporation N10/ICH 7 Family USB UHCI Controller #1 (rev 02)
00:1d.1 USB Controller: Intel Corporation N10/ICH 7 Family USB UHCI Controller #2 (rev 02)
00:1d.2 USB Controller: Intel Corporation N10/ICH 7 Family USB UHCI Controller #3 (rev 02)
00:1d.3 USB Controller: Intel Corporation N10/ICH 7 Family USB UHCI Controller #4 (rev 02)
00:1d.7 USB Controller: Intel Corporation N10/ICH 7 Family USB2 EHCI Controller (rev 02)
00:1e.0 PCI bridge: Intel Corporation 82801 Mobile PCI Bridge (rev e2)
00:1f.0 ISA bridge: Intel Corporation 82801GBM (ICH7-M) LPC Interface Bridge (rev 02)
00:1f.1 IDE interface: Intel Corporation 82801G (ICH7 Family) IDE Controller (rev 02)
00:1f.3 SMBus: Intel Corporation N10/ICH 7 Family SMBus Controller (rev 02)
02:00.0 System peripheral: JMicron Technology Corp. SD/MMC Host Controller
02:00.2 SD Host controller: JMicron Technology Corp. Standard SD Host Controller
02:00.3 System peripheral: JMicron Technology Corp. MS Host Controller
03:00.0 Network controller: Broadcom Corporation BCM4312 802.11b/g LP-PHY (rev 01)
04:00.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL8101E/RTL8102E PCI Express Fast Ethernet controller (rev 02)

[Vadim Peretokin (vperetokin)]

Upstream problem: http://www.phoronix.com/scan.php?page=article&item=ubuntu_natty_power&num=1

[amano (jyaku)]

I wonder why Phoronix doesn't simply share the commit that caused that regression.

[Nick Read (nickread)]

> Powertop shows 500-700 wake ups per second...

I'm only seeing 100-200 wake ups on Natty, similar to Maverick, but I'm still seeing getting *really* poor battery life. I also attemped disabling the proprietry FGLRX driver to reduce wake ups further (< 100), but this did not appear to give me any futher battery life.

> I wonder why Phoronix doesn't simply share the commit that caused that regression.

AFAICT, they're still bisecting to find it.

[michaellarabel (michael-michaellarabel)]

@amano: The PTS stack knows what branch pull caused the regression, but not the specific commit within there... Once I am certain of the commit within there that's causing the power issues, it will be shared. It's just not going as fast as I would like since the auto-bisecting is being done on a Core Duo.

[Pete Graner (pgraner)]

This is a known problem with the upstream linux kernel. More info can be found here:

http://www.phoronix.com/scan.php?page=article&item=ubuntu_natty_power&num=1

One this is resloved upstream ubuntu will get it thru stable updates or a backport.

[Leann Ogasawara (leannogasawara)]

I added some release notes text to the bug description. Thanks.

[stecklum (stecklum)]

I am running linux_2.6.38.orig.tar.gz + linux_2.6.38-8.41.diff.gz with reiser4-for-2.6.38, 2.6.38.2-breaks-suspend-to-disk, and v4l-dvb-as102 patches on my XPS M1330 with Ubuntu 8.10 (see lspci config below), and do _not_ encounter those problems. Powertop shows 60...100 wake ups per second as usual when idle.

00:00.0 Host bridge: Intel Corporation Mobile PM965/GM965/GL960 Memory Controller Hub (rev 0c)
00:01.0 PCI bridge: Intel Corporation Mobile PM965/GM965/GL960 PCI Express Root Port (rev 0c)
00:1a.0 USB Controller: Intel Corporation 82801H (ICH8 Family) USB UHCI Controller #4 (rev 02)
00:1a.1 USB Controller: Intel Corporation 82801H (ICH8 Family) USB UHCI Controller #5 (rev 02)
00:1a.7 USB Controller: Intel Corporation 82801H (ICH8 Family) USB2 EHCI Controller #2 (rev 02)
00:1b.0 Audio device: Intel Corporation 82801H (ICH8 Family) HD Audio Controller (rev 02)
00:1c.0 PCI bridge: Intel Corporation 82801H (ICH8 Family) PCI Express Port 1 (rev 02)
00:1c.1 PCI bridge: Intel Corporation 82801H (ICH8 Family) PCI Express Port 2 (rev 02)
00:1c.3 PCI bridge: Intel Corporation 82801H (ICH8 Family) PCI Express Port 4 (rev 02)
00:1c.5 PCI bridge: Intel Corporation 82801H (ICH8 Family) PCI Express Port 6 (rev 02)
00:1d.0 USB Controller: Intel Corporation 82801H (ICH8 Family) USB UHCI Controller #1 (rev 02)
00:1d.1 USB Controller: Intel Corporation 82801H (ICH8 Family) USB UHCI Controller #2 (rev 02)
00:1d.2 USB Controller: Intel Corporation 82801H (ICH8 Family) USB UHCI Controller #3 (rev 02)
00:1d.7 USB Controller: Intel Corporation 82801H (ICH8 Family) USB2 EHCI Controller #1 (rev 02)
00:1e.0 PCI bridge: Intel Corporation 82801 Mobile PCI Bridge (rev f2)
00:1f.0 ISA bridge: Intel Corporation 82801HEM (ICH8M) LPC Interface Controller (rev 02)
00:1f.1 IDE interface: Intel Corporation 82801HBM/HEM (ICH8M/ICH8M-E) IDE Controller (rev 02)
00:1f.2 SATA controller: Intel Corporation 82801HBM/HEM (ICH8M/ICH8M-E) SATA AHCI Controller (rev 02)
00:1f.3 SMBus: Intel Corporation 82801H (ICH8 Family) SMBus Controller (rev 02)
01:00.0 VGA compatible controller: nVidia Corporation GeForce 8400M GS (rev a1)
03:01.0 FireWire (IEEE 1394): Ricoh Co Ltd R5C832 IEEE 1394 Controller (rev 05)
03:01.1 SD Host controller: Ricoh Co Ltd R5C822 SD/SDIO/MMC/MS/MSPro Host Adapter (rev 22)
03:01.2 System peripheral: Ricoh Co Ltd R5C843 MMC Host Controller (rev 12)
03:01.3 System peripheral: Ricoh Co Ltd R5C592 Memory Stick Bus Host Adapter (rev 12)
03:01.4 System peripheral: Ricoh Co Ltd xD-Picture Card Controller (rev 12)
09:00.0 Ethernet controller: Broadcom Corporation NetLink BCM5906M Fast Ethernet PCI Express (rev 02)
0c:00.0 Network controller: Intel Corporation PRO/Wireless 4965 AG or AGN [Kedron] Network Connection (rev 61)
0d:00.0 SATA controller: JMicron Technologies, Inc. JMicron 20360/20363 AHCI Controller (rev 03)
0d:00.1 IDE interface: JMicron Technologies, Inc. JMicron 20360/20363 AHCI Controller (rev 03)

[Maddy (nmadhava)]

Same issue here after upgrading to Natty.
Seeing 350 - 550 wakeups per second, however the top cause seems
to be the "extra timer interrupt"

Backup reduced from 3 hours to 1hr 55 minutes now.

Wakeups-from-idle per second : 499.2 interval: 3.0s
no ACPI power usage estimate available

Top causes for wakeups:
  48.3% (303.3) [extra timer interrupt]
  18.6% (117.0) chrome
  12.1% ( 76.3) [iwlagn] <interrupt>
   4.9% ( 30.7) [ahci] <interrupt>

--
Maddy

[Sascha K. B. (saschakb)]

This regression is not only in Ubuntu. I'm running Archlinux - and for me - the regression takes also part in the Archkernel 2.6.38.x and the Liquorix Kernel 2.6.38.x.
The wakeups per second raise up to 1297/second, mostly via ehci_hcd:usb1... ehci_hcd:usbx, brcm80211 and ahci interrupts, keyboard, mousepad, touchpad interrupt.
The Turbo mode (2,31 GHz on my machine) uses up to 42.9% in idle mode, while except for 1% the rest goes to my 800 MHz powersaving mode. Even when being only in tty - and doing nothing, the cpu shows me that it uses 20% (instead of 1% in 2.6.32).
The battery of my Compaq CQ56 has only 2 hours on Kernel 2.6.38/2.6.39, while it has 3 hours on Kernel 2.6.32.x

With kernel 2.6.32.39 I'm down on 67 wakeups per second.

[stecklum (stecklum)]

To quantify my statement (posting #9) these are results of a 5min powertop dump in GDM with firefox+thunderbird idling for 2.6.36-1

Cn Avg residency
C0 (cpu running) ( 2.5%)
polling 0.0ms ( 0.0%)
C1 mwait 0.0ms ( 0.0%)
C2 mwait 0.3ms ( 0.5%)
C4 mwait 13.9ms (97.0%)
P-states (frequencies)
Turbo Mode 1.0%
  2.50 Ghz 0.0%
  1.60 Ghz 0.2%
  1200 Mhz 0.1%
   800 Mhz 98.6%
Wakeups-from-idle per second : 91.4 interval: 300.0s
no ACPI power usage estimate available
Top causes for wakeups:
  16.7% ( 10.8) kworker/0:0
  16.3% ( 10.5) [ata_piix] <interrupt>
  14.3% ( 9.2) [kernel scheduler] Load balancing tick
  12.4% ( 8.0) apt-check
   6.9% ( 4.4) [kernel core] hrtimer_start (tick_sched_timer)
   3.2% ( 2.0) cpufire_applet
   3.1% ( 2.0) [nvidia] <interrupt>
   3.0% ( 1.9) firefox-bin
...

and 2.6.38-8
Cn Avg residency
C0 (cpu running) ( 2.8%)
polling 0.0ms ( 0.0%)
C1 mwait 0.0ms ( 0.0%)
C2 mwait 0.2ms ( 0.3%)
C4 mwait 11.8ms (96.9%)
P-states (frequencies)
Turbo Mode 1.2%
  2.50 Ghz 0.1%
  1.60 Ghz 0.2%
  1200 Mhz 0.2%
   800 Mhz 98.3%
Wakeups-from-idle per second : 97.5 interval: 300.0s
no ACPI power usage estimate available
Top causes for wakeups:
  33.1% ( 25.4) kworker/0:0
  13.9% ( 10.7) [kernel scheduler] Load balancing tick
  13.9% ( 10.6) [extra timer interrupt]
   8.5% ( 6.5) [ata_piix] <interrupt>
   4.4% ( 3.3) [ahci, firewire_ohci] <interrupt>
   2.9% ( 2.2) cpufire_applet
   2.6% ( 2.0) firefox-bin
   2.6% ( 2.0) [nvidia] <interrupt>
   2.1% ( 1.6) gnome-terminal

The wakeup rate for 2.6.38 is higher indeed but not as excessive as seen by others.

[Sascha K. B. (saschakb)]

To be more specifig on my posting under #11, this is an output from powertop after 5 minutes idling

<Detaillierte C-Status Informationen sP-States (Frequenzen)
2,31 GHz 31,7%
1,71 GHz 3,7%
   800 MHz 64,6%

Aufwachen pro Sekunde : 617,8 Intervall: 10,0s
Keine ACPI Stromverbrauch-Schätzung verfügbar

Häufigste Ursachen für das Aufwachen:
  28,9% (118,3) USB Gerät 5-1 : i.Beat censo (TrekStor )
  24,9% (101,8) [ohci_hcd:usb4, ohci_hcd:usb5, ohci_hcd:usb6, ohci_hcd:usb7, fglrx[0]@PCI:1:5:0] <interrupt>
  24,0% ( 98,2) gtk-gnash
  11,4% ( 46,5) [ehci_hcd:usb1, ehci_hcd:usb2, ehci_hcd:usb3, brcm80211] <interrupt>
   5,0% ( 20,4) firefox
1,3% ( 5,3) [ahci] <interrupt>
1,2% ( 5,0) mc
0,5% ( 2,1) wget
0,5% ( 2,0) grep
0,4% ( 1,7) USB Gerät 5-2 : Microsoft IntelliMouse® Optical (Microsoft)
0,3% ( 1,2) kworker/0:1
0,3% ( 1,1) [kernel core] sk_reset_timer (tcp_delack_timer)
0,2% ( 1,0) ntpd
0,2% ( 1,0) tor
0,1% ( 0,5) udisks-daemon
0,1% ( 0,5) dropbox
0,1% ( 0,5) conky
0,1% ( 0,4) chromium
0,0% ( 0,2) kswapd0
0,0% ( 0,2) wicd-monitor
0,0% ( 0,2) [kernel core] run_timer_softirq (sync_supers_timer_fn)
0,0% ( 0,2) init
0,0% ( 0,2) wmfs
0,0% ( 0,1) [kernel core] run_timer_softirq (peer_check_expire)
0,0% ( 0,1) awk
0,0% ( 0,1) [kernel core] inet_twdr_hangman (inet_twdr_hangman)
0,0% ( 0,1) gconfd-2
0,0% ( 0,1) crond
0,0% ( 0,1) btrfs-endio-0
0,0% ( 0,1) [kernel core] bfq_completed_request (bfq_idle_slice_timer)
0,0% ( 0,1) khugepaged
0,0% ( 0,1) kworker/0:0
0,0% ( 0,1) wpa_supplicant
0,0% ( 0,1) flush-btrfs-1

[Sascha K. B. (saschakb)]

... and that is the maximum ... taken some minutes later.

<Detaillierte C-Status Informationen sP-States (Frequenzen)
2,31 GHz 35,4%
1,71 GHz 0,0%
   800 MHz 64,6%

Aufwachen pro Sekunde : 200000,0 Intervall: 0,0s
Keine ACPI Stromverbrauch-Schätzung verfügbar

[Luis Silva (lacsilva)]

Part of this problem may be related to Bug #771963.

[michaellarabel (michael-michaellarabel)]

Bug #771963 may be an indirect contributor to bad power performance, but in terms of the dramatic power set-backs, I (and others) have encountered it on Intel and NVIDIA hardware as well. My tracking of the issue still seems to be leading back to something going awry in mm.

[Ilja Sekler (ilja-sekler-)]

Out of curiosity, I let my Asus Eee PC 1000H triple boot Ubuntu 11.04 - Xubuntu 11.04 - Ubuntu 10.04.2 (camera, wifi and bluetooth disabled, no network connection) idle exactly for an hour on battery with locked screen (no screensaver). Battery energy drain on Ubuntu Natty (classic session): 4,4 Wh (average consumption 4,4 W). Battery energy drain on Ubuntu 10.04.2: 7,7 Wh (7,7 W on average). Big advantage for Natty so far.

[pablomme (pablomme)]

I don't think the duplicate status is correct. As far as we know this is a regression between 2.6.37 and 2.6.38, not something present in 2.6.32.

[MountainX (dave-mountain)]

I hope this is an appropriate question here:

Can someone explain this comment?

>Leann Ogasawara on 2011-04-25
>Changed in ubuntu-release-notes:
>status: In Progress → Fix Released

The status says fix released, yet it doesn't appear to be fixed. Does that mean the fix is coming or does it mean the fix didn't really fix it?

Thanks for any reply.

[Julian Kalinowski (julakali)]

According to phoronix benchmarks (which appears to be down right now), its a regression starting in 2.6.34.
The results or at least a quick summary is posted in the bug: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/524281/comments/189

Since the other bug report exists for more than a year now, and neither wakeups nor power consumption have been reduced in kernels newer than 2.6.34, i think this bug is a duplicate of #524281

[Nikolai Prokoschenko (nikolai)]

@MountainX: it means the part of this bug relevant to the release notes has been fixed, i.e. it's been mentioned there as known problem.

[Shimi Chen (shimi-chen)]

@MountainX It means that an entry was made in the release notes for Natty
about this regression.

On 28 April 2011 22:17, MountainX <email address hidden> wrote:

> I hope this is an appropriate question here:
>
> Can someone explain this comment?
>
> >Leann Ogasawara on 2011-04-25
> >Changed in ubuntu-release-notes:
> >status: In Progress → Fix Released
>
> The status says fix released, yet it doesn't appear to be fixed. Does
> that mean the fix is coming or does it mean the fix didn't really fix
> it?
>
> Thanks for any reply.
>
> --
> You received this bug notification because you are a direct subscriber
> of the bug.
> https://bugs.launchpad.net/bugs/760131
>
> Title:
> Power consumption raised significantly in natty
>
> Status in Release Notes for Ubuntu:
> Fix Released
> Status in “linux” package in Ubuntu:
> Confirmed
> Status in “linux” source package in Natty:
> Confirmed
> Status in “linux” source package in Oneiric:
> Confirmed
>
> Bug description:
> This is a regression moving from maverick->natty. With the same sort
> of light interactive usage, Battery life has gone from ~4 hours to
> 2:45 or so.
>
> Running powertop shows the system in P-state "Turbo mode" 10-15% of
> the time. Typically this was almost zero in maverick, 98+% of the
> time maverick was in the lowest state.
>
> wakeups reported in the 5-600/s range where previously (from memory)
> it was 200 ish. i915 driver is high on the list. This is
> irrespective of whether running unity or classic desktop.
>
> ProblemType: Bug
> DistroRelease: Ubuntu 11.04
> Package: linux-image-2.6.38-8-generic 2.6.38-8.42
> Regression: Yes
> Reproducible: Yes
> ProcVersionSignature: Ubuntu 2.6.38-8.42-generic 2.6.38.2
> Uname: Linux 2.6.38-8-generic x86_64
> AlsaVersion: Advanced Linux Sound Architecture Driver Version 1.0.23.
> Architecture: amd64
> ArecordDevices:
> **** List of CAPTURE Hardware Devices ****
> card 0: Intel [HDA Intel], device 0: CONEXANT Analog [CONEXANT Analog]
> Subdevices: 1/1
> Subdevice #0: subdevice #0
> AudioDevicesInUse:
> USER PID ACCESS COMMAND
> /dev/snd/controlC0: james 2068 F.... pulseaudio
> CRDA: Error: [Errno 2] No such file or directory
> Card0.Amixer.info:
> Card hw:0 'Intel'/'HDA Intel at 0xf2620000 irq 47'
> Mixer name : 'Conexant CX20561 (Hermosa)'
> Components : 'HDA:14f15051,17aa20ff,00100000'
> Controls : 16
> Simple ctrls : 8
> Card29.Amixer.info:
> Card hw:29 'ThinkPadEC'/'ThinkPad Console Audio Control at EC reg 0x30,
> fw 7XHT24WW-1.06'
> Mixer name : 'ThinkPad EC 7XHT24WW-1.06'
> Components : ''
> Controls : 1
> Simple ctrls : 1
> Card29.Amixer.values:
> Simple mixer control 'Console',0
> Capabilities: pswitch pswitch-joined penum
> Playback channels: Mono
> Mono: Playback [on]
> Date: Wed Apr 13 15:03:35 2011
> EcryptfsInUse: Yes
> HibernationDevice: RESUME=UUID=bb479652-f524-4abe-b1b0-27646d6deebc
> InstallationMedia: Ubuntu 10.10 "Maverick Meerkat" - Release Candidate
> amd64 (20100928)
> MachineType: LENOVO 7465CTO
> ProcEnviron:
> LANGUAGE=en_US:en_GB:en
> PATH=(custom, user)
> LANG=en_US....

[pablomme (pablomme)]

@MountainX: that task refers to the release notes (i.e., this bug should be mentioned in the release notes). It is (see https://wiki.ubuntu.com/NattyNarwhal/ReleaseNotes [known issues > kernel section]), so that is done. The bug itself is not fixed.

@Julian: it would seem then that there have been two independent power regressions, very likely caused by different code changes, and they should be kept as separate bugs since they should be solved separately

[reini (rrumberger)]

@Julian: From this bug's description, it's a regression from maverick, which comes with 2.6.35. In the comment to bug #524281 you reference, the big discrepancy is between 34 and 35, i.e. it should have been apparent in maverick and not just have appeared in natty.
IMHO this all points to different bugs with similar symptoms.

[tippettm (tippettm)]

Phoronix is up again.

http://www.phoronix.com/scan.php?page=article&item=linux_kernel_regress2&num=4 and raw results at http://openbenchmarking.org/result/1104256-GR-RADEONPOW94

As you can see from the graph, there are two regressions for a lot of the workloads. This is shown by the three clusters of consistent results. The first is between 2.6.34 and 2.6.35, and the second is between 2.6.38.

Note that this is looking at the kernel as the variant on an Ubuntu 8.04 base install. It is highly likely that there is a confluence of other factors that may have exacerbated the impact with Natty.

[Andrea Grandi (andreagrandi)]

I've upgraded from Ubuntu 10.10 to Ubuntu 11.04 on my Asus EeePC 1005PE netbook and I've tried running powertop.
I've 130-160 wakeups-from-idle per second. Is it a good value? So this bug is not affecting all notebook/netbook?

[Matthew Hessel (matt-hessel)]

built latest kernel 2.6.39-rc5 from latest sources and powertop shows normal values for power utilization again.

processors show 99 percent C6 mwait residency, full charge battery estimate is up to 3 hours compared to 50 percent C6 mwait and 1.2 hours off a full charge

[Ilja Sekler (ilja-sekler-)]

> I've 130-160 wakeups-from-idle per second. Is it a good value?

Probably not, because with the very same hardware on Maverick (kernel 2.6.35-29-generic) powertop counts only 35 interrupts/sec with wireless disabled and about 85 with wireless enabled. Estimated power consumption is 6,6 vs 6,5 W.

On Natty (amd64) with Nvidia binary blob, running on Asus M4N68T motherboard with AMD Athlon II X2 240e CPU, powertop reports about 185 interrupts per second when idle. Disabling remote control polling in dvb-usb module brings this value down to 35 interrupts/sec.

It means, without a carefully defined and unified hardware and software configuration, bare figures are meaningless.

[Jamie Lokier (jamie-shareable)]

Comment #24, reini wrote on 2011-04-28:
> Julian: From this bug's description, it's a regression from maverick, which comes with 2.6.35.
> In the comment to bug #524281 you reference, the big discrepancy is between 34 and 35,
> i.e. it should have been apparent in maverick and not just have appeared in natty.
> IMHO this all points to different bugs with similar symptom

Not necessarily. In another area (touchscreen drivers) I recently found to my great surprise that Ubuntu's 2.6.35-28 kernel does not contain all changes in the 2.6.35 vanilla upstream kernel!

That was a shock, I expected 2.6.35-28 to be based on 2.6.35 with Ubuntu changes or stability backports.

I don't know if that applies to this bug, but don't assume every upstream change between 34 and 35 is necessarily in Maverick's 2.6.35 kernel.

[James Ferguson (jamesf)]

The i915 driver appears to be creating approx 95000 interrupts per second -

$ cat /proc/irq/45/spurious
count 99516
unhandled 369
last_unhandled 12764990 ms

$ ls /proc/irq/45/
affinity_hint i915@pci:0000:00:02.0 node smp_affinity spurious

[Ed Guenter (edgue)]

Just a side note: a coworker of mine is using a Lenovo X200 ... and he tells me: since he updated to 2.3.38 (on top of 10.10) ... his battery life time increased ...

[Matthew Hessel (matt-hessel)]

looks like part of this was noticed back in September.

http://lkml.org/lkml/2010/9/30/100

I took the Natty kernel sources and default config and manually applied the diffs from that thread.

the load balance thread that keeps waking the kernel drops significantly when idle. - went from 300 or so wakeups per second minmum on my Core 2 duo laptop to about 30 at minimum.

I'm no expert, but from reading the thread, it looks like the tickless patches in the last few kernels combined with the SMP code causes excessive ticking in the kernel when there is no work to do, so it can't effectively idle at lower sleep states.

I haven't tested this, but single CPU/single core machines might not be affected by this, as there is no function for the load balancing code when there is only one core.

[nanog (sorenimpey)]

Looks like the discussion at lkml has been going on for a long time.

http://lkml.org/lkml/2010/4/26/249

This history of this set of patches correlates well with the timeline of increased power consumption from phoronix testing. IMO, the loss of power consumption does not equal the benefit of increased kernel responsiveness especially for laptops.

>I haven't tested this, but single CPU/single core machines might not be affected by this,
As one would expect this bug is present on hyperthreaded single core atom chips.
For those who want increased battery life on their atom chips...2.6.34 is running terrifically in natty.

[Matthew Hessel (matt-hessel)]

It might be a cludge, but a workaround might be to disable the
tickless kernel option, multiple core systems and hyperthreading
would still work with a slight increase in overhead- at least until
this has an accepted patch from the kernel folks.

On 5/12/11, nanog <email address hidden> wrote:
> Looks like the discussion at lkml has been going on for a long time.
>
> http://lkml.org/lkml/2010/4/26/249
>
> This history of this set of patches correlates well with the timeline of
> increased power consumption from phoronix testing. IMO, the loss of
> power consumption does not equal the benefit of increased kernel
> responsiveness especially for laptops.
>
>>I haven't tested this, but single CPU/single core machines might not be
>> affected by this,
> As one would expect this bug is present on hyperthreaded single core atom
> chips.
> For those who want increased battery life on their atom chips...2.6.34 is
> running terrifically in natty.
>
> --
> You received this bug notification because you are a direct subscriber
> of the bug.
> https://bugs.launchpad.net/bugs/760131
>
> Title:
> Power consumption raised significantly in natty
>
> Status in Release Notes for Ubuntu:
> Fix Released
> Status in “linux” package in Ubuntu:
> Confirmed
> Status in “linux” source package in Natty:
> Confirmed
> Status in “linux” source package in Oneiric:
> Confirmed
>
> Bug description:
> This is a regression moving from maverick->natty. With the same sort
> of light interactive usage, Battery life has gone from ~4 hours to
> 2:45 or so.
>
> Running powertop shows the system in P-state "Turbo mode" 10-15% of
> the time. Typically this was almost zero in maverick, 98+% of the
> time maverick was in the lowest state.
>
> wakeups reported in the 5-600/s range where previously (from memory)
> it was 200 ish. i915 driver is high on the list. This is
> irrespective of whether running unity or classic desktop.
>
> ProblemType: Bug
> DistroRelease: Ubuntu 11.04
> Package: linux-image-2.6.38-8-generic 2.6.38-8.42
> Regression: Yes
> Reproducible: Yes
> ProcVersionSignature: Ubuntu 2.6.38-8.42-generic 2.6.38.2
> Uname: Linux 2.6.38-8-generic x86_64
> AlsaVersion: Advanced Linux Sound Architecture Driver Version 1.0.23.
> Architecture: amd64
> ArecordDevices:
>  **** List of CAPTURE Hardware Devices ****
>  card 0: Intel [HDA Intel], device 0: CONEXANT Analog [CONEXANT Analog]
>    Subdevices: 1/1
>    Subdevice #0: subdevice #0
> AudioDevicesInUse:
>  USER PID ACCESS COMMAND
>  /dev/snd/controlC0: james 2068 F.... pulseaudio
> CRDA: Error: [Errno 2] No such file or directory
> Card0.Amixer.info:
>  Card hw:0 'Intel'/'HDA Intel at 0xf2620000 irq 47'
>    Mixer name : 'Conexant CX20561 (Hermosa)'
>    Components : 'HDA:14f15051,17aa20ff,00100000'
>    Controls : 16
>    Simple ctrls : 8
> Card29.Amixer.info:
>  Card hw:29 'ThinkPadEC'/'ThinkPad Console Audio Control at EC reg 0x30,
> fw 7XHT24WW-1.06'
>    Mixer name : 'ThinkPad EC 7XHT24WW-1.06'
>    Components : ''
>    Controls : 1
>    Simple ctrls : 1
> Card29.Amixer.values:
>  Simple mixer c...

[Jim Collins (jamtrino)]

Well, when comparing to Win7, I'm seeing a 25% increase in battery drain (and lots of heat) when using kernel 2.6.32 with Lucid. Not sure what to make of this.

With Natty, I'm seeing a 30% increase in battery drain (and lots of heat).

[Miroslav (dzundam-deactivatedaccount)]

@nanog

It seems to be not the cause...

See:
http://www.phoronix.com/scan.php?page=news_item&px=OTQ1Ng

[Matthew Hessel (matt-hessel)]

Yes, I noticed.

I compiled up a lucid kernel - 2.6.34, runs fine on natty. Strange
thing I see is that the ACPI reported power usage is the same on my
laptop, about 19 - 21 watts, yet I get 3.5 hours on that kernel
instead of the 2.1 hours on the natty kernel or the 2.6.39 source from
the linus tree. I'm using the same nvidia blob on both.

If I switch to nouveau it increases the power usage noticeably, adds
about 4 watts in both cases while running gnome or unity and the xorg
edger's mesa and xorg binaries.

I don't understand how the ACPI power usage can be the same and still
have such a large variance for run time on the battery...

On 5/20/11, Miroslav <email address hidden> wrote:
> @nanog
>
> It seems to be not the cause...
>
> See:
> http://www.phoronix.com/scan.php?page=news_item&px=OTQ1Ng
>
> --
> You received this bug notification because you are a direct subscriber
> of the bug.
> https://bugs.launchpad.net/bugs/760131
>
> Title:
> Power consumption raised significantly in natty
>
> Status in Release Notes for Ubuntu:
> Fix Released
> Status in “linux” package in Ubuntu:
> Confirmed
> Status in “linux” source package in Natty:
> Confirmed
> Status in “linux” source package in Oneiric:
> Confirmed
>
> Bug description:
> This is a regression moving from maverick->natty. With the same sort
> of light interactive usage, Battery life has gone from ~4 hours to
> 2:45 or so.
>
> Running powertop shows the system in P-state "Turbo mode" 10-15% of
> the time. Typically this was almost zero in maverick, 98+% of the
> time maverick was in the lowest state.
>
> wakeups reported in the 5-600/s range where previously (from memory)
> it was 200 ish. i915 driver is high on the list. This is
> irrespective of whether running unity or classic desktop.
>
> ProblemType: Bug
> DistroRelease: Ubuntu 11.04
> Package: linux-image-2.6.38-8-generic 2.6.38-8.42
> Regression: Yes
> Reproducible: Yes
> ProcVersionSignature: Ubuntu 2.6.38-8.42-generic 2.6.38.2
> Uname: Linux 2.6.38-8-generic x86_64
> AlsaVersion: Advanced Linux Sound Architecture Driver Version 1.0.23.
> Architecture: amd64
> ArecordDevices:
>  **** List of CAPTURE Hardware Devices ****
>  card 0: Intel [HDA Intel], device 0: CONEXANT Analog [CONEXANT Analog]
>    Subdevices: 1/1
>    Subdevice #0: subdevice #0
> AudioDevicesInUse:
>  USER PID ACCESS COMMAND
>  /dev/snd/controlC0: james 2068 F.... pulseaudio
> CRDA: Error: [Errno 2] No such file or directory
> Card0.Amixer.info:
>  Card hw:0 'Intel'/'HDA Intel at 0xf2620000 irq 47'
>    Mixer name : 'Conexant CX20561 (Hermosa)'
>    Components : 'HDA:14f15051,17aa20ff,00100000'
>    Controls : 16
>    Simple ctrls : 8
> Card29.Amixer.info:
>  Card hw:29 'ThinkPadEC'/'ThinkPad Console Audio Control at EC reg 0x30,
> fw 7XHT24WW-1.06'
>    Mixer name : 'ThinkPad EC 7XHT24WW-1.06'
>    Components : ''
>    Controls : 1
>    Simple ctrls : 1
> Card29.Amixer.values:
>  Simple mixer control 'Console',0
>    Capabilities: pswitch pswitch-joined penum
>    Playback channels: Mono
>    Mono: Playback [on]
> Date: We...

[Christopher Patrick (cpatrick08)]

do you know how much longer till the fix is in natty-updates

[Philip Muškovac (yofel)]

@cpatrick08: I believe there is no fix for this issue yet. Not even the mainline kernel is fixed yet since the issue seems to be pretty hard to track down.

[z06gal (z06gal)]

Any news on this issue? Alot of folks are experiencing real heat issues. Hope there is a fix soon

[lljccoael (lljccoael-deactivatedaccount)]

And many common laptops are still plagued by the lack of workable default fan controls, of which most newcomers are completely unaware. What a double whammy.

[Salih EMIN (salih-emin)]

This is the reason why there should not be a Generic kernel. This kernel in contrast of server kernel, is the same for Desktops, Laptops and Netbooks.
The server kernel is optimized for server purposes, whereases the generic kernel is the same for Desktop and Mobile PC's (Laptops/Netbooks). Unfortunately the Desktop PC is not the same with the Mobile PC's. Desktops don't have consumption problems but Mobile PC's do have.

We can not demand the generic kernel to behave in the same way in Mobile PC's as it behaves in Desktop PC's. That is why personally, after a fresh install of a new Ubuntu version, I recompile the generic kernel with options that are best for my Laptop:

-- Conservative Governance (default is performance that drains my battery and produces unnecessary heat )
-- 1000hz cycle (if I recall correctly default is 250Hz, which creates more wakeups for CPU)
-- Core2 Family (optimized for my i7core CPU, to remove any unnecessary generic code from kernel )

The above are some of the ways that I have accomplished better battery life and less heat on my Laptop.
IMO, there kernel team should pack Ubuntu kernel in 2 flavors (as they did for server edition):
--Desktop PC optimized options Kernel
--Mobile PC optimized options Kernel

[Scott Kitterman (kitterman)]

On Sunday, May 29, 2011 09:19:02 PM you wrote:
> This is the reason why there should not be a Generic kernel. ...

No. It's completely unrelated. It's a bug, but we don't have a fix yet.
Please rant elsewhere.

[Salih EMIN (salih-emin)]

@Scott
I am really sorry if my comment and suggestions sounded like a rant or trolling. I never or will ever do that sort of commenting. I do have this bug (as I am using the same stock Ubuntu kernel) but with the options that I previously stated, I had reduced the overall impact of that bug in my Laptops battery/heat levels.
I hope and I am confident that this bug will sooner or later be fixed and will be pushed in backport repos.

Again, my intentions are to give some personal observations and some useful workarounds that I use for years in my Mobile PC's and not to rant, so I am truly sorry if my words were misunderstood.

[nanog (sorenimpey)]

@miroslav,
so those lkml threads were a different issue.
i really hope michael larabel finds time to perform those bisects!

i'd report this upstream myself but without git bisects i am way too chicken to face the ire of testy kernel devs.
_______________________________________________________________

for those who are still affected by this bug i recommend trying 2.6.34 from here:

http://kernel.ubuntu.com/~kernel-ppa/mainline/v2.6.34.7-maverick/

this takes care of both major power regressions noted by michael larabel. my netbook battery (6 cell) life increased from ~3.7 hrs on mainline 2.6.38 to 6.5 hrs on 2.6.34!

[lljccoael (lljccoael-deactivatedaccount)]

^ Thanks for the suggestion, nanog.

@Scott: Salih wasn't ranting.

[Philip Muškovac (yofel)]

No need to go back as far as 2.6.34. The regression is in 2.6.38 so going back to 2.6.37 is enough.

@salih:
-- Conservative Governance (default is performance that drains my battery and produces unnecessary heat )
It's set to performance so the system boots as fast as possible. /etc/init.d/ondemand sets it to ondemand later which in most cases is better than conservative.

[z06gal (z06gal)]

My wireless driver is bundled with the natty kernel so I'll keep monitoring my temps till there is a fix. You would think it would be a high priority for the number of people dealing with this.

[Mike (mike-fdb)]

Has anyone tried powertop2? It differs from powertop (gets data from kernel's perf infrastructure) and might give a hint on power burner.

[Uwe (gandalf.the.grey)]

I'm running a Dell Inspiron 6400 with Karmic and could not find any significant difference between the output of Powertop on Karmic and Natty (started from CD). Between about 40 and 200 wakeups per second were reported each and the CPU were running more than 90 % of the time on the lowest frequenzy of 800 MHz.

[Peter Sasi (peter-sasi)]

I have updated my tests for both the latest 2.6.38 kernel of natty and the latest 2.6.34 kernel from mainline as suggested, on my Thinkpad T61 laptop.
After login I have sudoed previously, then run:
sudo powertop -d -t 60 > ~/Desktop/powertop_dump-`uname -r`.log
Running on battery, not having started anything but one terminal window after logon.
Both kernels, results below: there is a visible difference still! 15,5W versus 17,9W meaning 3,4 hours versus 2,9 hours = a half an hour of time on battery!

I think both power regressions found by Phoronix at 2.6.35 and 2.6.38 are there.
See http://www.phoronix.com/scan.php?page=article&item=linux_kernel_regress2

2.6.34:
PowerTOP 1.13 (C) 2007 - 2010 Intel Corporation

Collecting data for 60 seconds

Cn Avg residency
C0 (cpu running) ( 1,9%)
C0 0,0ms ( 0,0%)
C1 mwait 0,0ms ( 0,0%)
C2 mwait 0,5ms ( 0,4%)
C6 mwait 6,2ms (97,7%)
P-states (frequencies)
Turbo Mode 0,9%
  2,50 Ghz 0,0%
  1,60 Ghz 0,0%
  1200 Mhz 0,0%
   800 Mhz 99,1%
Wakeups-from-idle per second : 164,1 interval: 60,0s
Power usage (ACPI estimate): 15,5W (3,4 hours)
Top causes for wakeups:
  29,5% ( 61,0) [uhci_hcd:usb5, yenta, nvidia] <interrupt>
  24,2% ( 50,0) [kernel core] hdaps_mousedev_poll (hdaps_mousedev_poll)
  19,3% ( 39,9) [kernel core] hrtimer_start (tick_sched_timer)
  13,4% ( 27,8) [kernel scheduler] Load balancing tick
   4,8% ( 9,9) gwibber-service
   2,4% ( 5,0) [ata_piix] <interrupt>
   1,7% ( 3,6) compiz
   1,1% ( 2,2) nautilus
   0,8% ( 1,7) gnome-terminal

2.6.38:
PowerTOP 1.13 (C) 2007 - 2010 Intel Corporation

Collecting data for 60 seconds

Cn Avg residency
C0 (cpu running) ( 2,2%)
polling 0,1ms ( 0,0%)
C1 mwait 0,0ms ( 0,0%)
C2 mwait 0,6ms ( 0,7%)
C6 mwait 4,5ms (97,1%)
P-states (frequencies)
Turbo Mode 0,9%
  2,50 Ghz 0,0%
  2,00 Ghz 0,0%
  1,60 Ghz 0,1%
   800 Mhz 99,0%
Disk accesses:
The application 'gvfsd-metadata' is writing to file 'home-3c698ca6.log' on /dev/sda5
The application 'gvfsd-metadata' is writing to file 'home-3c698ca6.log' on /dev/sda5
The application 'rs:main Q:Reg' is writing to file 'auth.log' on /dev/sda5
The application 'rs:main Q:Reg' is writing to file 'auth.log' on /dev/sda5
The application 'rs:main Q:Reg' is writing to file 'auth.log' on /dev/sda5
The application 'gvfsd-metadata' is writing to file 'home.SF5MWV' on /dev/sda5
The application 'gvfsd-metadata' is writing to file 'home.SF5MWV' on /dev/sda5
Wakeups-from-idle per second : 227,9 interval: 60,0s
Power usage (ACPI estimate): 17,9W (2,9 hours)
Top causes for wakeups:
  24,9% ( 61,0) [uhci_hcd:usb5, yenta, nvidia] <interrupt>
  20,4% ( 50,0) [kernel core] hdaps_mousedev_poll (hdaps_mousedev_poll)
  15,1% ( 36,9) [extra timer interrupt]
  14,8% ( 36,4) [kernel core] hrtimer_start (tick_sched_timer)
   7,4% ( 18,2) compiz
   6,2% ( 15,2) [kernel scheduler] Load balancing tick
   4,1% ( 10,0) gwibber-service
   2,2% ( 5,5) kworker/0:0
   1,6% ( 4,0) [ata_piix] <interrupt>
   0,0% ( 0,0)D gvfsd-metadata
   0,0% ( 0,0)D rs:main Q:Reg
   0,9% ( 2,3) nautilus
   0,7% ( 1,7) gnome-termina...

[lljccoael (lljccoael-deactivatedaccount)]

I can confirm that using kernel 2.6.34, as suggested, solves the problem on my hardware. I have not tried 2.6.37 yet.

[Jonas Wagner (wagner-j)]

I have tried 2.6.34 and 2.6.32, they both consume 4 Watts (~20%) less than 2.6.38.

[Jonas Wagner (wagner-j)]

I get the following on a Thinkpad T60 with 2.6.37 (from http://kernel.ubuntu.com/~kernel-ppa/mainline/v2.6.37.6-natty/):

Cn Avg residency
C0 (cpu running) ( 9.1%)
polling 0.2ms ( 0.0%)
C1 halt 0.2ms ( 0.0%)
C2 0.5ms ( 1.3%)
C3 4.1ms (89.6%)
P-states (frequencies)
  2.17 Ghz 5.4%
  1.67 Ghz 0.2%
  1333 Mhz 0.4%
  1000 Mhz 94.1%
Wakeups-from-idle per second : 244.7 interval: 300.0s
Power usage (ACPI estimate): 19.6W (0.3 hours)
Top causes for wakeups:
  46.8% (189.5) [Rescheduling interrupts] <kernel IPI>
  14.8% ( 60.1) [radeon] <interrupt>
  11.7% ( 47.4) kworker/0:0
  11.5% ( 46.4) [kernel scheduler] Load balancing tick
   4.9% ( 19.7) desktopcouch-se

and the stock 2.6.38:

Cn Avg residency
C0 (cpu running) ( 9.6%)
polling 2.9ms ( 0.0%)
C1 halt 0.2ms ( 0.0%)
C2 0.3ms ( 1.0%)
C3 3.2ms (89.4%)
P-states (frequencies)
  2.17 Ghz 5.8%
  1.67 Ghz 0.1%
  1333 Mhz 0.6%
  1000 Mhz 93.5%
Wakeups-from-idle per second : 306.2 interval: 300.0s
Power usage (ACPI estimate): 24.0W (0.4 hours)
Top causes for wakeups:
  37.6% (166.6) [Rescheduling interrupts] <kernel IPI>
  13.6% ( 60.1) [radeon] <interrupt>
  11.8% ( 52.0) [kernel scheduler] Load balancing tick
  11.5% ( 50.9) kworker/0:0
  10.3% ( 45.4) [extra timer interrupt]

[Adrian Wechner (adrian-wechner)]

this article is quiet interesting. investigating the possible issues with a time frame when the changes were made which gave the power consumption increase. maybe helpful:
http://www.phoronix.com/scan.php?page=news_item&px=OTM3NQ

[Andy Whitcroft (apw)]

On Mon, May 30, 2011 at 01:19:02AM -0000, Salih EMIN wrote:

> -- Conservative Governance (default is performance that drains my battery and produces unnecessary heat )

The default is performance for boot speed, this gets switched to
Conservative after login and remains that way we do not leave you in
Performance mode unless there is another bug.

> -- 1000hz cycle (if I recall correctly default is 250Hz, which creates more wakeups for CPU)

This is likely to increase not decrease your wakeups, as you have made
timers more granular and scheduling ticks more common. It is not clear
how this improves anything, in our testing it cost you 10% of additional
CPU to perform tasks which should equate to more heat.

> -- Core2 Family (optimized for my i7core CPU, to remove any unnecessary generic code from kernel )

Fair enough.

-apw

[z06gal (z06gal)]

I saw a post on the phoronix forums and decided to install the 2.6.39.1 kernel. It has solved my power issues greatly. I have been running it all afternoon and my temps have remained low and my battery life is back up to normal. Here is the post:

"I just stumbled across a patch for (soon to arrive) version 2.6.39.1:

http://thread.gmane.org/gmane.linux....02/focus=12505

Maybe this small patch solves the power regression."

[S. Christian Collins (s-chriscollins)]

For those who are wondering, the post mentioned in z06gal's post can be found here:
http://phoronix.com/forums/showthread.php?55062-Nailing-Down-The-Linux-Kernel-Power-Regressions/page3

[lljccoael (lljccoael-deactivatedaccount)]

I would just remind the curious, however, that installing any vanilla kernel will render Remastersys useless, as it depends on Ubuntu kernel patches like Aufs. As such, LiveCDs made with the 2.6.39.1 kernel, for example, will not be able to boot. Since remasters are so popular in these situations, it is necessary to either wait for an official fix, or downgrade to an Ubuntu patched kernel that doesn't have the problem.

[Andy Whitcroft (apw)]

The patch which is being discussed seems to be this one:

  commit 7467571f4480b273007517b26297c07154c73924
  Author: Tero Kristo <email address hidden>
  Date: Thu Feb 24 17:19:23 2011 +0200

    cpuidle: menu: fixed wrapping timers at 4.294 seconds

    Cpuidle menu governor is using u32 as a temporary datatype for storing
    nanosecond values which wrap around at 4.294 seconds. This causes errors
    in predicted sleep times resulting in higher than should be C state
    selection and increased power consumption. This also breaks cpuidle
    state residency statistics.

    cc: <email address hidden> # .32.x through .39.x
    Signed-off-by: Tero Kristo <email address hidden>
    Signed-off-by: Len Brown <email address hidden>

This patch seems to be in 2.6.38.8 and 2.6.39.1 stable updates which should be included in official kernels in the next few weeks.

[Andy Whitcroft (apw)]

I have pulled in just the fix in comment #60 into some test kernel for Natty. If those of you who are seeing this issue could test those kernel and report whether they help with battery life that would be helpful. Kernels are at the URL below:

    http://people.canonical.com/~apw/lp760131-natty/

Please report any testing back here. Thanks.

[clubber (fkramer)]

I installed the kernel and will now test it during the day. Report will follow

uname -a
Linux kassiopeia 2.6.38-9-generic #43lp760131v201106060906 SMP Mon Jun 6 08:27:00 UTC 2011 x86_64 x86_64 x86_64 GNU/Linux

[Alessandro Lazzari (lazzari-alessandro)]

Testing it on a Sony Vaio VGN-SR21M

Wakeups from idle per second have generally dropped from 150 - 200 to a 50 - 100 range.

Linux VGN-SR21M 2.6.38-9-generic #43lp760131v201106060906 SMP Mon Jun 6 08:27:00 UTC 2011 x86_64 x86_64 x86_64 GNU/Linux

Cn Avg residency P-states (frequenze)
C0 (cpu occupata) ( 1,1%) Modalità turbo 0,1%
polling 0,0ms ( 0,0%) 2,27 Ghz 0,0%
C1 mwait 0,1ms ( 0,0%) 1,60 Ghz 0,0%
C2 mwait 0,1ms ( 0,0%) 800 Mhz 99,9%
C4 mwait 11,5ms (98,9%)

Wakeups da idle per secondo : 86,7 interval: 15,0s
Energia utilizzata (stima ACPI): 15,5W (1,1 ore) (long term: 18,8W,/0,9h)

Cause principali di wakeup:
  34,5% ( 36,9) [extra timer interrupt]
  18,5% ( 19,7) desktopcouch-se
  13,6% ( 14,5) [radeon] <interrupt>
   9,4% ( 10,0) kworker/0:0
   0,0% ( 0,0)D rs:main Q:Reg

[Simon Strandman (nejsimon)]

The patch hopefully improves the situation but that bug is probably present in maverick's kernel 2.6.35 as well so there might be other issues too.

[Jonas Wagner (wagner-j)]

I still see the same power consumption as with 2.6.38-8 (see #54).
Wakeups dropped by ~40.

Linux wm 2.6.38-9-generic #43lp760131v201106060906 SMP Mon Jun 6 08:27:00 UTC 2011 i686 i686 i386 GNU/Linux

Cn Avg residency
C0 (cpu running) ( 9.1%)
polling 0.8ms ( 0.0%)
C1 halt 0.0ms ( 0.0%)
C2 0.3ms ( 0.8%)
C3 3.8ms (90.2%)
P-states (frequencies)
  2.17 Ghz 4.5%
  1.67 Ghz 0.2%
  1333 Mhz 0.1%
  1000 Mhz 95.2%
Wakeups-from-idle per second : 264.0 interval: 300.0s
Power usage (ACPI estimate): 23.9W (1.1 hours)
Top causes for wakeups:
  36.9% (143.3) [Rescheduling interrupts] <kernel IPI>
  15.5% ( 60.1) [radeon] <interrupt>
  13.0% ( 50.6) [kernel scheduler] Load balancing tick
  12.8% ( 49.9) kworker/0:0
   7.4% ( 28.7) [extra timer interrupt]

[z06gal (z06gal)]

My laptop continues to run better on the 2.6.39.1 kernel

Wakeups-from-idle per second : 32.4 interval: 15.0s
Power usage (ACPI estimate): 2.3W (3.5 hours) (long term: 2.3W,/3.4h)

Top causes for wakeups:
  33.9% ( 44.3) chromium-browse
  27.1% ( 35.4) docky
  16.9% ( 22.1) [iwlagn] <interrupt>
   6.3% ( 8.2) kworker/0:1
   2.6% ( 3.4) kworker/0:0
   2.0% ( 2.7) [ahci] <interrupt>
   2.0% ( 2.7) [i915] <interrupt>
   2.0% ( 2.6) [kernel scheduler] Load balancing tick
   1.6% ( 2.1) Xorg

[Alessandro Lazzari (lazzari-alessandro)]

Well, just tried to install Karmic on the same Vaio VGN-SR21M, and here are the results.

Wakeups in the range of 20 - 30, while I thought that the 120 - 150 range was the norm.
It seems to me that the consumption level rised with Maverick, since I also tested a Lucid server that stays around 20 - 30 all the time.

Results are for the Vaio laptop.

Linux ubuntu 2.6.31-14-generic #48-Ubuntu SMP Fri Oct 16 14:04:26 UTC 2009 i686 GNU/Linux

Cn Avg residency P-states (frequencies)
C0 (cpu running) ( 0.0%) 2.27 Ghz 0.3%
polling 0.0ms ( 0.0%) 2.27 Ghz 0.1%
C1 mwait 0.0ms ( 0.0%) 1.60 Ghz 0.0%
C2 mwait 0.0ms ( 0.0%) 800 Mhz 99.6%
C3 mwait 47.0ms (100.0%)

Wakeups-from-idle per second : 21.4 interval: 15.0s
no ACPI power usage estimate available

Top causes for wakeups:
  23.1% ( 6.3) <kernel core> : hrtimer_start (tick_sched_timer)
  16.3% ( 4.5) <kernel core> : hrtimer_start_range_ns (tick_sched_timer)
  14.6% ( 4.0) <kernel core> : usb_hcd_poll_rh_status (rh_timer_func)
  12.9% ( 3.5) <interrupt> : extra timer interrupt
  11.9% ( 3.3) <interrupt> : ahci
   4.9% ( 1.3) <kernel core> : neigh_periodic_timer (neigh_periodic_timer)

[Laurens Bosscher (laurens-laurensbosscher)]

Installing it right now, Before results are :

Dell XPS15 with Intel Sandybridge (screen on lowest brightness)

Wakeups : 200-300
Power usage : 13,7 - 16,8 W

Cn Avg residency P-states (frequencies)
C0 (cpu running) (14,9%) Turbo Mode 3,5%
polling 0,3ms ( 0,0%) 1400 Mhz 0,3%
C1 mwait 0,4ms ( 2,5%) 1200 Mhz 0,2%
C2 mwait 0,8ms ( 1,4%) 1000 Mhz 0,7%
C3 mwait 0,9ms ( 0,2%) 800 Mhz 95,1%
C4 mwait 4,2ms (81,0%)

I will update in a few minutes with results with the patch.

[Jorge Eduardo (jorge-birck)]

Hi Andy and everyone!

Testing it right now

Linux birck 2.6.38-9-generic #43lp760131v201106060906 SMP Mon Jun 6 08:08:37 UTC 2011 i686 GNU/Linux

Seems ok! I will update again. Intel i3

[Jorge Eduardo (jorge-birck)]

Fixed for me ( intel i3 )v:

http://people.canonical.com/~apw/lp760131-natty/

Thank you Andy!

[Jorge Eduardo (jorge-birck)]

I still have high idle cpu usage seemingly caused by kworker processes. Any ideia how to fix? intel i3

[Michael Gratton (mjog)]

While 2.6.38-9-generic seems to have reduced wakeups on my MacbookPro 8,1 slightly, I'm still getting 500-600/s from Interprocessor interrupts and 100-200/s from i915 PCI-MSI-edge interrupts.

[lljccoael (lljccoael-deactivatedaccount)]

Not fixed for this laptop (a Toshiba Satellite L305-S5937, Intel GM45). It certainly runs just as hot with this patched 2.6.38-9 as it did with the stock 2.6.38, sadly, and the battery still discharges at an alarming rate. Only reverting to the 2.6.34 kernel returns this computer to normalcy so far. Reverting to 2.6.37 is not enough.

[z06gal (z06gal)]

@cjcolella Have you tried the 2.6.39.1 kernel? I have no battery or temp issues with this kernel and the performance is excellent.

[Alessandro Lazzari (lazzari-alessandro)]

@Mike Gratton

Video card drivers generate a VBLANK interrupt when using opengl, i.e. for compiz, as explained on the Powertop website:

http://www.lesswatts.org/projects/powertop/known.php

If you log in the 2D desktop, where this interrupt is not used, you'll see a good decrease in wake-ups.
Ubuntu one sync daemons also generate a good amount of interrupts and, along with the opengl issue of above, they seem to contribute to the general increase in consumption noticed after Lucid.

However, if you experience 500 - 600/s just from interprocessor interrupts, there's probably a kernel issue at work here, but I believe that there's also an added increase in consumption due to the overall changes in the various OS's.

Anyway, on Natty stock 2.6.38 kernel, switching to 2D desktop and stopping Ubuntu one daemons, shows a decrease in wakeups from the 120 - 150 range to 40 - 60, on my Vaio VGN-SR21M.

[Alessandro Lazzari (lazzari-alessandro)]

Some details about what posted above.

30,1% ( 68,2) [radeon] <interrupt>
 29,9% ( 67,6) [extra timer interrupt]
 8,7% ( 19,6) desktopcouch-se
 8,0% ( 18,0) compiz

[clubber (fkramer)]

I couldn't measure an improvement with the changes made to the kernel provided by Andy in #61
but since i installed kernel 2.6.38-10 at least my CPU is not literally burning anymore with wakeups remaining at arround 300 on a Lenovo x300

[Matthew Hessel (matt-hessel)]

@Alex - is the radeon driver the Xorg driver?

Using the proprietary fglrx one will probably improve things quite a bit.

Power consumption on the open source stack for both amd and nvidia is
still quite high.

My friend's eepc 1015T dropped about 4-5 watts when I installed the
proprietary driver.

On 6/9/11, Alessandro Lazzari <email address hidden> wrote:
> Some details about what posted above.
>
> 30,1% ( 68,2) [radeon] <interrupt>
> 29,9% ( 67,6) [extra timer interrupt]
> 8,7% ( 19,6) desktopcouch-se
> 8,0% ( 18,0) compiz
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/760131
>
> Title:
> Power consumption raised significantly in natty
>
> Status in Release Notes for Ubuntu:
> Fix Released
> Status in “linux” package in Ubuntu:
> Triaged
> Status in “linux” source package in Natty:
> Incomplete
> Status in “linux” source package in Oneiric:
> Triaged
>
> Bug description:
> This is a regression moving from maverick->natty. With the same sort
> of light interactive usage, Battery life has gone from ~4 hours to
> 2:45 or so.
>
> Running powertop shows the system in P-state "Turbo mode" 10-15% of
> the time. Typically this was almost zero in maverick, 98+% of the
> time maverick was in the lowest state.
>
> wakeups reported in the 5-600/s range where previously (from memory)
> it was 200 ish. i915 driver is high on the list. This is
> irrespective of whether running unity or classic desktop.
>
> ProblemType: Bug
> DistroRelease: Ubuntu 11.04
> Package: linux-image-2.6.38-8-generic 2.6.38-8.42
> Regression: Yes
> Reproducible: Yes
> ProcVersionSignature: Ubuntu 2.6.38-8.42-generic 2.6.38.2
> Uname: Linux 2.6.38-8-generic x86_64
> AlsaVersion: Advanced Linux Sound Architecture Driver Version 1.0.23.
> Architecture: amd64
> ArecordDevices:
>  **** List of CAPTURE Hardware Devices ****
>  card 0: Intel [HDA Intel], device 0: CONEXANT Analog [CONEXANT Analog]
>    Subdevices: 1/1
>    Subdevice #0: subdevice #0
> AudioDevicesInUse:
>  USER PID ACCESS COMMAND
>  /dev/snd/controlC0: james 2068 F.... pulseaudio
> CRDA: Error: [Errno 2] No such file or directory
> Card0.Amixer.info:
>  Card hw:0 'Intel'/'HDA Intel at 0xf2620000 irq 47'
>    Mixer name : 'Conexant CX20561 (Hermosa)'
>    Components : 'HDA:14f15051,17aa20ff,00100000'
>    Controls : 16
>    Simple ctrls : 8
> Card29.Amixer.info:
>  Card hw:29 'ThinkPadEC'/'ThinkPad Console Audio Control at EC reg 0x30,
> fw 7XHT24WW-1.06'
>    Mixer name : 'ThinkPad EC 7XHT24WW-1.06'
>    Components : ''
>    Controls : 1
>    Simple ctrls : 1
> Card29.Amixer.values:
>  Simple mixer control 'Console',0
>    Capabilities: pswitch pswitch-joined penum
>    Playback channels: Mono
>    Mono: Playback [on]
> Date: Wed Apr 13 15:03:35 2011
> EcryptfsInUse: Yes
> HibernationDevice: RESUME=UUID=bb479652-f524-4abe-b1b0-27646d6deebc
> InstallationMedia: Ubuntu 10.10 "Maverick Meerkat" - Release Candidate
> amd64 (20100928)
> MachineType: LENOVO 7465CTO
> ProcEnviron:
>  LANGUAGE=...

[Jorge Eduardo (jorge-birck)]

Andy's kernel doesnt fix my friend's laptop burning. Dell Inspiron 11z

[michaellarabel (michael-michaellarabel)]

FYI, the 2.6.39.1 "fix" seems to just be an improvement but not a correction for the actual regressions - http://www.phoronix.com/vr.php?view=16104

[Alessandro Lazzari (lazzari-alessandro)]

@Matthew Hessel

Yes, the [radeon] driver is the xorg one. Intel drivers show that behaviour too, even if on the powertop website they state that the issue is now fixed.

Here are the figures for a Dell D630 ATG laptop, Natty stock kernel. The i915 driver causes 62+ wakeups /sec. with unity/gnome effects, and none in the 2D desktop.

21,5% ( 81,7) [extra timer interrupt]
16,5% ( 62,7) [i915] <interrupt>
16,0% ( 60,7) [ehci_hcd:usb2, uhci_hcd:usb3, uhci_hcd:usb5] <interrupt>
16,0% ( 60,7) Dispositivo USB 5-2 : Compact Optical Mouse 500 (Microsoft )
5,9% ( 22,3) kworker/0:

[lljccoael (lljccoael-deactivatedaccount)]

@z06gal: Thank you for the suggestion, but unfortunately, 2.6.39.1 shows little improvement on this hardware (Toshiba laptop L305-S5937, Core 2 Duo Intel processors, integrated Intel GM45, etc.).

I would add, however, that I am using Kubuntu now, and have the same issue in Ubuntu. Even without compositing in each, the problem is still manifest.

Here is a summary of Powertop using kernel 2.6.37 and Kwin effects:

Wakeups-from-idle per second : 1122.8 interval: 15.0s
no ACPI power usage estimate available

Top causes for wakeups:
  22.6% ( 77.1) [i915] <interrupt>
  19.8% ( 67.5) USB device 5-1 : USB OPTICAL MOUSE (PIXART)
  15.0% ( 51.0) [ehci_hcd:usb2, uhci_hcd:usb5] <interrupt>
   5.9% ( 20.0) knotify4
   5.0% ( 17.0) kworker/0:0
   4.4% ( 15.0) USB device 1-4 : USB2.0-CRW (Generic)
   4.2% ( 14.4) [ehci_hcd:usb1, uhci_hcd:usb6] <interrupt>
   3.4% ( 11.7) [ath9k] <interrupt>
   0.1% ( 0.4)D plasma-desktop
   2.8% ( 9.5) kworker/u:0
   2.8% ( 9.4) [kernel scheduler] Load balancing tick

Now, i915 is the top cause here, though the "Load balancing tick" often tops at different intervals. Disabling desktop effects doesn't change the results much.

So, disabling desktop effects and reverting to the 2.6.34 kernel results in the following:

Wakeups-from-idle per second : 1182.8 interval: 3.0s
no ACPI power usage estimate available

Top causes for wakeups:
  21.1% ( 44.3) [kernel scheduler] Load balancing tick
  10.5% ( 22.0) [ehci_hcd:usb2, uhci_hcd:usb5] <interrupt>
   9.5% ( 20.0) knotify4
   8.7% ( 18.3) USB device 5-1 : USB OPTICAL MOUSE (PIXART)
   8.6% ( 18.0) [ehci_hcd:usb1, uhci_hcd:usb6] <interrupt>
   7.9% ( 16.7) USB device 1-4 : USB2.0-CRW (Generic)
   4.8% ( 10.0) [kernel core] ath_ani_calibrate (ath_ani_calibrate)
   4.6% ( 9.7) [ath9k] <interrupt>
   4.3% ( 9.0) [kernel core] hrtimer_start (tick_sched_timer)
   2.9% ( 6.0) [ahci] <interrupt>
   2.1% ( 4.3) phy0
   1.9% ( 4.0) [kernel core] usb_hcd_poll_rh_status (rh_timer_func)
   1.9% ( 4.0) kwin
   1.3% ( 2.7) [hda_intel] <interrupt>
   1.3% ( 2.7) alsa-sink
   1.3% ( 2.7) threaded-ml
   1.3% ( 2.7) USB device 2-3 : My Passport 070A (Western Digital)

So I found this other bug report about "[kernel scheduler] Load balancing tick" on Core 2 Duo. I wonder, as Mr. Hessel mentioned back in May, could this be a contributing factor?

https://bugs.launchpad.net/debian/+source/linux-2.6/+bug/524281

[Sergio Zanchetta (primes2h)]

Great, battery life increased of about 2 hours for me.

Asus 1015PEM, here are specs:
http://laptop.qa.ubuntu.com/qatracker/laptop/Natty/28925

[Matthew Hessel (matt-hessel)]

Cjcolla - with over 1000 wakeups per second, I would not think there
is much chance of the processor entering a sleep state at all.

It looks to me like all of the wakeups are from your usb devices, were
you doing much with the mouse or anything like that while taking the
snapshot?

Earlier in the thread is a link for 2.6.38.9 kernel patched to fix the
kernel wakeup timer wrapping issue, on my laptop it doesn't seem to
change the maximum power draw in powertop, but when wakeup events are
triggered, the power draw doesn't seem to spike as bad.

With desktop effects on and I am running unity with the proprietary
nvidia driver, I idle at about 21-22 watts and about 100 wakeups per
second.

With gnome (no effects). That drops to around 20 watts.

Nouveau will kick it up to 26 watts at idle.

With the lucid kernel running on natty, I get idle down to about 18 -
19 watts and the nvidia driver. 24 watts with nouveau.

You might want to check into installing laptop-mode-tools and
verifying you have cpufrequtils installed.

My friend's eepc 1015 was upgraded to natty, and did not get those
installed by default, I added and enabled those and installed the
proprietary fglrx driver to drop him down from 16 watts to about 9 at
idle with unity running full effects.

I was running conky configured with conky-colors (mentioned at
webupd8) and found that the 1 second polling was eating up a lot of
power. Top wakeups were from the window manager and the kernel load
balancing process. Killing conky dropped both of those down again.

I changed conky to only poll every 5 seconds, now the draw isn't too
bad, maybe 1-2 watts additional.

On 6/9/11, cjcolella <email address hidden> wrote:
> @z06gal: Thank you for the suggestion, but unfortunately, 2.6.39.1 shows
> little improvement on this hardware (Toshiba laptop L305-S5937, Core 2
> Duo Intel processors, integrated Intel GM45, etc.).
>
> I would add, however, that I am using Kubuntu now, and have the same
> issue in Ubuntu. Even without compositing in each, the problem is still
> manifest.
>
> Here is a summary of Powertop using kernel 2.6.37 and Kwin effects:
>
> Wakeups-from-idle per second : 1122.8 interval: 15.0s
>
> no ACPI power usage estimate available
>
> Top causes for wakeups:
> 22.6% ( 77.1) [i915] <interrupt>
> 19.8% ( 67.5) USB device 5-1 : USB OPTICAL MOUSE (PIXART)
> 15.0% ( 51.0) [ehci_hcd:usb2, uhci_hcd:usb5] <interrupt>
> 5.9% ( 20.0) knotify4
> 5.0% ( 17.0) kworker/0:0
> 4.4% ( 15.0) USB device 1-4 : USB2.0-CRW (Generic)
> 4.2% ( 14.4) [ehci_hcd:usb1, uhci_hcd:usb6] <interrupt>
> 3.4% ( 11.7) [ath9k] <interrupt>
> 0.1% ( 0.4)D plasma-desktop
> 2.8% ( 9.5) kworker/u:0
> 2.8% ( 9.4) [kernel scheduler] Load balancing tick
>
>
> Now, i915 is the top cause here, though the "Load balancing tick" often tops
> at different intervals. Disabling desktop effects doesn't change the results
> much.
>
> So, disabling desktop effects and reverting to the 2.6.34 kernel results
> in the following:
>
> Wakeups-from-idle per second : 1182.8 interval: 3.0s
>
> no ACPI power usage estimate available
>
> Top causes for wakeups:
> 21.1% ( 44.3)...

[lljccoael (lljccoael-deactivatedaccount)]

No, I wasn't doing anything at all with the mouse when I recorded the information from Powertop. Also, the other USB device was just an empty flash card that I left plugged in. I closed out of everything before refreshing Powertop, including all file manager windows. Perhaps not surprisingly, leaving any browser open brings the wakeups far above 2000 per second, consistently.

Out of curiosity, I unplugged the flash drive and mouse, and the results are the same. i915 and Load balancing tick are, as usual, at the top of the list, but why PS/2 keyboard/mouse/touchpad (or more often "USB device 1-4 : USB2.0-CRW (Generic)" at about the same value) should, consistently, especially when I'm not using them, seems weird. I don't know.

    PowerTOP version 1.13 (C) 2007 Intel Corporation

Cn Avg residency P-states (frequencies)
C0 (cpu running) ( 1.8%) 2.17 Ghz 0.0%
polling 0.0ms ( 0.0%) 1.67 Ghz 0.0%
C1 mwait 10.0ms (49.0%) 1333 Mhz 0.0%
C3 mwait 0.5ms (49.3%) 1000 Mhz 100.0%

Wakeups-from-idle per second : 1141.2 interval: 3.0s
no ACPI power usage estimate available

Top causes for wakeups:
  19.7% ( 33.7) [kernel scheduler] Load balancing tick
  16.8% ( 28.7) [i915] <interrupt>
  16.2% ( 27.7) PS/2 keyboard/mouse/touchpad interrupt
   8.4% ( 14.3) phy0

I'll try the stuff you mentioned.

[Julian Kalinowski (julakali)]

This is getting OT, but how can you get 1141 wakups/sec when your top wakeup cause has only 33.7 wakeup/s ?
I recently booted windows xp, which only consumes 8.5W compared to 12W in idle on ubuntu :(
I'd really like the powersaving issue to be adressed..

[Matthew Hessel (matt-hessel)]

I did some digging on the lkml forums (hard to add links from the bby)

It looks like all of them relate to a series of commits for the kernel
IPI process. It looks to be added as a way to load balance processes
accross multiple cores in the fair scheduler (CFS) when the kernel is
running tickless (config_no_hz).

IPI is supposed to evaluate what core should handle the load balancing
process across all cores at any given time, based on which ones are
busy and idle. It looks to me to prefer keeping the load balance
process on a lightly loaded core so that the heavier processes don't
get impacted by the additional work of managing the other cores'
processes.

It sounds to me that it is a good way to keep high performance in a
server (especially as you increase the number of cores in the system,)
but I wonder if that may be the wong approach for a smaller system
looking to maximize power efficiency instead of raw performance.

On a single socket dual core system like atom, brazos, core2duo etc,
it doesn't seem to be necessary as the cores on a single socket share
cache and other resources.

I wonder if it would make more sense in the smaller systems (single
socket) to assign the load balancing process on one core, skip the IPI
altogether and only wake the other cores when there is work to do.
Perhaps even prefer to run new processes itself until it needs to wake
another core.

(Unfortunately, I am not a kernel hacker - If I am misunderstanding
any of this, I would welcome someone to correct me :)

On 6/10/11, Julian Kalinowski <email address hidden> wrote:
> This is getting OT, but how can you get 1141 wakups/sec when your top wakeup
> cause has only 33.7 wakeup/s ?
> I recently booted windows xp, which only consumes 8.5W compared to 12W in
> idle on ubuntu :(
> I'd really like the powersaving issue to be adressed..
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/760131
>
> Title:
> Power consumption raised significantly in natty
>
> Status in Release Notes for Ubuntu:
> Fix Released
> Status in “linux” package in Ubuntu:
> Triaged
> Status in “linux” source package in Natty:
> Incomplete
> Status in “linux” source package in Oneiric:
> Triaged
>
> Bug description:
> This is a regression moving from maverick->natty. With the same sort
> of light interactive usage, Battery life has gone from ~4 hours to
> 2:45 or so.
>
> Running powertop shows the system in P-state "Turbo mode" 10-15% of
> the time. Typically this was almost zero in maverick, 98+% of the
> time maverick was in the lowest state.
>
> wakeups reported in the 5-600/s range where previously (from memory)
> it was 200 ish. i915 driver is high on the list. This is
> irrespective of whether running unity or classic desktop.
>
> ProblemType: Bug
> DistroRelease: Ubuntu 11.04
> Package: linux-image-2.6.38-8-generic 2.6.38-8.42
> Regression: Yes
> Reproducible: Yes
> ProcVersionSignature: Ubuntu 2.6.38-8.42-generic 2.6.38.2
> Uname: Linux 2.6.38-8-generic x86_64
> AlsaVersion: Advanced Linux Sound Architecture Driver Version 1.0.23.
> Architecture: amd64
> ...

[lljccoael (lljccoael-deactivatedaccount)]

@ Mr. Kalinowski: Yeah, sorry to go off-topic. I have no idea why the Powertop results seem so bizarre, but it looks like everyone here got high wakeup-from-idle totals.

I'll just note that after testing many kernels and different patches from 2.6.34 to 2.6.39.1, the latest Maverick 2.6.35-28-generic kernel performs the best on my hardware. I have no idea why. I also installed and activated cpufrequtils, as Mr. Hessel suggested, and ran 'sudo cpufreq-set -g conservative'.

[Alan Jenkins (aj504)]

*Grar*. The "conservative" regulator only saves power on really ancient hardware. On current hardware it *loses* you battery life.

C.f. this post by Matthew Garret who hacks kernels for Redhat and is generally to credit for a lot of laptop-specific testing, work on platform-specific ACPI drivers, etc.

<http://mjg59.livejournal.com/88608.html?thread=1059616>

You should also be able to find a similar explanation somwhere from Arjan, who worked (works?) for *intel*.

If I remember correctly, current kernels will even select "conservative" automatically on the ancient CPUs where it is appropriate.

Believe it or not, Linux distributions aren't _quite_ so incompetent as to leave massive battery life/heat wins on the table for years, by not bothering to simply set an existing configuration option when running on a laptop.

While I'm ranting, I'd like to point out this bug is about a regression - something that got _worse_ on Linux, that in the worst case can be tracked down by looking through older versions. ("bisection"). It won't help us to compare Windows - and try and reverse-engineer what it's doing better - when we know that there's a version of _Linux_ that's doing better. You'll only be able to isolate a windows-linux difference if you know you're _not_ suffering from this regression, _and_ you're running a recent kernel (otherwise you may be overlooking more recent power improvements. Runtime-PM works very nicely on my EEEPC, although it's not enabled by default at the moment). It's worth mentioning in a general discussion, but it's not going to help fix bug #760131.

[Tobias Stegmann (punischdude)]

Using a Lenovo U160 (Core i5-520M) the power consumption improves slightly with the patched 2.6.38-9 kernel. Wakeups per second stay the same, though.

[Diane Trout (diane-trout)]

Thinkpad X61 Tablet Core 2 Duo L7500 @ 1.6 GHz

about the same 280 wakeups per second between 2.3.38-8 and 2.3.38-9 from linux-image-2.6.38-9-generic_2.6.38-9.43lp760131v201106060906_amd64.deb. however ACPI estimated power usage goes from ~19W to 17.4W

I also tried 2.3.32-33-generic from the ubuntu pool, and got 18.4W with 320 wakeups per second.

In all cases I started power top, enable everything it suggested and wait for gwibber to finish syncing after boot. I had emacs, firefox (open to simple pages), and gnome-terminal open, plus whatever starts by default in 11.04. Once top showed the system mostly idle I pulled the AC to get the ACPI power usage estimate.

[Filiprino (filiprino)]

Hello,
I'm suffering too from the shortening of the battery's life. Powertop reports the following data:

     PowerTOP version 1.13 (C) 2007 Intel Corporation

Cn Residencia media Estados P (frecuencias)
C0 (cpu ejecutando) ( 9,8%) 1,67 Ghz 11,3%
sondeando 28,6ms ( 0,2% 1333 Mhz 2,0%
C1 mwait 0,2ms ( 0,4%) 1000 Mhz 86,7%
C2 mwait 0,8ms (11,9%)
C4 mwait 2,1ms (77,7%)

Despertares por segundo: 536,2 intervalo: 15,0s
Consumo de energía (estimado por ACPI): 10,0W (0,3 horas) (largo plazo: 6,5W,/0,5h)

Causas principales de despertares:
  26,9% (156,3) [tiempo de interrupción extra]
  16,7% ( 96,9) plugin-containe
  15,8% ( 91,6) PS/2 teclado/ratón/touchpad interrupción
  10,3% ( 60,1) [i915] <interrupción>
   8,0% ( 46,5) [iwlagn] <interrupción>
   5,0% ( 29,3) [planificador del núcleo] Tick del balanceo de carga
   5,0% ( 29,1) kworker/0:0
   0,1% ( 0,5)D upowerd
   3,0% ( 17,6) compiz
   2,3% ( 13,2) firefox-bin
   1,8% ( 10,6) [ahci] <interrupción>

"i915" and "extra timer interrupt" are always at the very top of the list. My computer is an ASUS EEEPC 1005PE with an Intel Wi-Fi Link 5300 card.
The kernel version I'm using is 2.6.38-8-generic #42-Ubuntu SMP

[lljccoael (lljccoael-deactivatedaccount)]

"*Grar*. The "conservative" regulator only saves power on really ancient hardware. On current hardware it *loses* you battery life."

Good to know, Mr. Jenkins. I suppose this information ought to be widely proliferated, as many Linux kernel manuals (as Google readily shows), published as recently as this year, still say that Conservative is better for all battery devices, and makes no distinction between old and new hardware. At any rate, it was just an experiment, since I'd guess the main issue for most is the excessive heat. At least, most people I've talked to with this problem thought that was the most immediately noticeable change on their laptops, and discovered the decreased battery life and found out about the increased power consumption regression later.

"Believe it or not, Linux distributions aren't _quite_ so incompetent as to leave massive battery life/heat wins on the table for years, by not bothering to simply set an existing configuration option when running on a laptop."

As far I understand what's been going on here, people are just trying to discover short-term treatments, not a cure, so to speak, while they patiently wait for a true fix from the Kernel Team.

[Salih EMIN (salih-emin)]

I run some PowerTop tests on 5 kernels (the full results are in the attached file) on Ubuntu 11.04 (Dell Studio 1557)
In summary:
-------------------------------------------------------------------------
Kernel 2.6.38-10 (from Pre-released repositories)

Wakeups-from-idle per second : 54.2 interval: 1.4s
Power usage (ACPI estimate): 1.4W (2.9 hours)
-------------------------------------------------------------------------
Kernel 2.6.38-9-generic [NO fglrx ATI driver] (the patched one mentioned here in launchpad at #70 )

Wakeups-from-idle per second : 39.5 interval: 3.0s
Power usage (ACPI estimate): 2.2W (1.8 hours)
-------------------------------------------------------------------------
Kernel 2.6.38-8-generic (The natty default linux kernel)

Wakeups-from-idle per second : 50.6 interval: 3.0s
Power usage (ACPI estimate): 2.2W (1.7 hours)
-------------------------------------------------------------------------
Kernel 2.6.34-020634-generic (from kernel ppa mainline - lucid)

Wakeups-from-idle per second : 57.7 interval: 15.0s
Power usage (ACPI estimate): 1.5W (1.9 hours) (long term: 1.6W,/1.7h)
------------------------------------------------------------------------
kernel 2.6.32-0206324118-generic (from kernel ppa mainline - lucid)

Wakeups-from-idle per second : 76.7 interval: 15.0s
Power usage (ACPI estimate): 4.1W (0.5 hours)
------------------------------------------------------------------------

[Maciej Borzecki (maciek-borzecki)]

I'm running a Thinkpad T60, T5600, X1300 and checked on Fedora 15. They have both new powertop (~1.98) and more options enabled in the kernel. Anyway, I can get ~17W power usage there (kernel 2.6.38.7) which is slightly more what I got on Fedora 14 (2.6.35.13). So the laptop has boasting 3:40 minutes of battery time. On Ubuntu 11.04 it's sadly 2:40.
Anyway, I guess it does make sense to compare the kernel configs at some point.

[Laurent Bonnaud (laurent-bonnaud)]

Here is a benchmark that confirms the problem:

http://www.tomshardware.com/reviews/ubuntu-11.04-natty-narwhal,2943-13.html

[lljccoael (lljccoael-deactivatedaccount)]

Any updates at all?

[Simon (nowis74)]

Michael Larabel from Phoronix seems to have found the source of the power regression for the kernel 2.6.38 and a workaround for it.

http://www.phoronix.com/scan.php?page=news_item&px=OTYwMQ

[Jaxon (bjackson0971)]

On page 3 of the Phoronix thread:

http://phoronix.com/forums/showthread.php?56437-Burning-Through-Power-Linux-Regressions-Found/page3

Michael says there is a fix as long as you don't have a bad BIOS. This statement makes me think it might be this patch that disables ASPM for some hardware:

PCI: Disable ASPM if BIOS asks us to
https://lkml.org/lkml/2010/12/2/303

I have a shell script to set a bunch of low power setting and it started printing an "Operation not permitted" error when it tried to:

echo powersave > /sys/module/pcie_aspm/parameters/policy

I didn't know the cause, but it was related to a kernel upgrade a while back.

Recently, I found this message in the boot messages:

kernel: pci 0000:08:00.0: disabling ASPM on pre-1.1 PCIe device. You can enable it with 'pcie_aspm=force'

I tried adding that flag to the boot parameters in my grub config and have had no problems for the last couple weeks. I cannot report any difference in battery life since I rarely run my laptop unplugged. There seems to be no difference in the number of wakeups reported by Powertop.

There was also an earlier patch to disable ASPM:

https://lkml.org/lkml/2011/3/4/246

If this is the cause of the increase in power use, then it is not really a bug or a regression. It is a deliberate change to fix a problem with certain hardware. It's easy enough for everyone to test out stability with the flag to force ASPM on.

[michaellarabel (michael-michaellarabel)]

Yes, the 2.6.38 power regression is caused by Active-State Power Management... http://www.phoronix.com/vr.php?view=16181 Any vendors with further questions are welcome to contact me at Michael [at] MichaelLarabel.com.

[Roberto Cássio Jr. (rcsdnj)]

Great work, Michael :)

Is there any simple way to check if my system is affected by this problem or not (i.e., querying ASPM status)?

[Reinhard (reinhard-fink)]

Thanks Michael,

for finding out.

On a HP 7 pavillion laptop, I tried your solution with "pcie_aspm=force", but is it possible, that aspm is switched off afterwards?
here my dmesg-lines:

153: [ 0.000000] Kernel command line: BOOT_IMAGE=/boot/vmlinuz-2.6.39-0-generic root=UUID=c.. ro pcie_aspm=force
154: [ 0.000000] PCIe ASPM is forcedly enabled
but later:
223: [ 0.805341] ACPI FADT declares the system doesn't support PCIe ASPM, so disable it
and
357: [ 1.232560] Unable to assume _OSC PCIe control. Disabling ASPM
and
365: [ 1.236368] pci0000:7f: Requesting ACPI _OSC control (0x1d)
366: [ 1.236371] Unable to assume _OSC PCIe control. Disabling ASPM

keep in training for "oktoberfest" (Less are not better)

[felixcorrales (felixcorrales-yahoo)]

MichaelLarabel

Can you please explain step by step how to configure the "pcie_aspm=force" boot command line?

"most people affected by this issue will want to add "pcie_aspm=force" to their boot command line"

"ASPM policies can also be set via /sys/module/pcie_aspm/parameters/policy"

[Carlos Moffat (carlos-eldiabloenlosdetalles)]

I can confirm comment #102 in my Lenovo x220. Forcing pcie_aspm doesn't work.

@FelixCorrales, take a look at /etc/default/grub, you can add that parameter to GRUB_CMDLINE_LINUX_DEFAULT

Cheers.

[michaellarabel (michael-michaellarabel)]

When forcing pcie_aspm and it says in the dmesg it's forced, even if it then says that it's disabled, it's still forced on -- their text output is confusing.

[Paulo J. S. Silva (pjssilva)]

Answering #102:

Edit /etc/default/grub as root and add "pcie_aspm=force" in the string GRUB_CMDLINE_LINUX_DEFAULT.
After that run "sudo update-grub" and reboot.

Answering #104, and completing #105:

If you find something like

[ 0.000000] PCIe ASPM is forcedly enabled

in dmesg output it should be enabled. You can disregard the next message

[ 0.775371] ACPI FADT declares the system doesn't support PCIe ASPM, so disable it

ASPM is enabled.

I am testing in my computer now and the option seems to make a real difference in may machine (Dell Latitude E6410).

[Reinhard (reinhard-fink)]

The easiest way to TEST Active-State Power Management is just to modify the command-line in the boot menu.
after
root=UUID=c.. ro ..
add
pcie_aspm=force.

then you can check in
/var/log/dmesg
after
Kernel command line: BOOT_IMAGE=/boot/vmlinuz-2.6.39...
if
"PCIe ASPM is forcedly enabled".

if it is working add in /etc/default/grub to GRUB_CMDLINE_LINUX_DEFAULT="other stuff pcie_aspm=force"

[Jonas Wagner (wagner-j)]

Adding 'pcie_aspm=force' decreased the power consumption by about 20% on a Thinkpad T60.
The wakeup count is still about the same.

[felixcorrales (felixcorrales-yahoo)]

The /etc/default/grub file have the following lines:

GRUB_CMDLINE_LINUX_DEFAULT=”quiet splash”
GRUB_CMDLINE_LINUX=”"

Where should I add "pcie_aspm=force"?

[Martin Spacek (mspacek)]

@felix, add it after the word "splash", like this:

GRUB_CMDLINE_LINUX_DEFAULT=”quiet splash pcie_aspm=force”

On 2011-06-27 09:59, felixcorrales wrote:
> The /etc/default/grub file have the following lines:
>
> GRUB_CMDLINE_LINUX_DEFAULT=”quiet splash”
> GRUB_CMDLINE_LINUX=”"
>
> Where should I add "pcie_aspm=force"?
>

[Sérgio Faria (sergio91pt)]

To default like this:
GRUB_CMDLINE_LINUX_DEFAULT=”quiet splash pcie_aspm=force”

2011/6/27 felixcorrales <email address hidden>

> The /etc/default/grub file have the following lines:
>
> GRUB_CMDLINE_LINUX_DEFAULT=”quiet splash”
> GRUB_CMDLINE_LINUX=”"
>
> Where should I add "pcie_aspm=force"?
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/760131
>
> Title:
> Power consumption raised significantly in natty
>
> Status in Release Notes for Ubuntu:
> Fix Released
> Status in “linux” package in Ubuntu:
> Triaged
> Status in “linux” source package in Natty:
> Incomplete
> Status in “linux” source package in Oneiric:
> Triaged
>
> Bug description:
> This is a regression moving from maverick->natty. With the same sort
> of light interactive usage, Battery life has gone from ~4 hours to
> 2:45 or so.
>
> Running powertop shows the system in P-state "Turbo mode" 10-15% of
> the time. Typically this was almost zero in maverick, 98+% of the
> time maverick was in the lowest state.
>
> wakeups reported in the 5-600/s range where previously (from memory)
> it was 200 ish. i915 driver is high on the list. This is
> irrespective of whether running unity or classic desktop.
>
> ProblemType: Bug
> DistroRelease: Ubuntu 11.04
> Package: linux-image-2.6.38-8-generic 2.6.38-8.42
> Regression: Yes
> Reproducible: Yes
> ProcVersionSignature: Ubuntu 2.6.38-8.42-generic 2.6.38.2
> Uname: Linux 2.6.38-8-generic x86_64
> AlsaVersion: Advanced Linux Sound Architecture Driver Version 1.0.23.
> Architecture: amd64
> ArecordDevices:
> **** List of CAPTURE Hardware Devices ****
> card 0: Intel [HDA Intel], device 0: CONEXANT Analog [CONEXANT Analog]
> Subdevices: 1/1
> Subdevice #0: subdevice #0
> AudioDevicesInUse:
> USER PID ACCESS COMMAND
> /dev/snd/controlC0: james 2068 F.... pulseaudio
> CRDA: Error: [Errno 2] No such file or directory
> Card0.Amixer.info:
> Card hw:0 'Intel'/'HDA Intel at 0xf2620000 irq 47'
> Mixer name : 'Conexant CX20561 (Hermosa)'
> Components : 'HDA:14f15051,17aa20ff,00100000'
> Controls : 16
> Simple ctrls : 8
> Card29.Amixer.info:
> Card hw:29 'ThinkPadEC'/'ThinkPad Console Audio Control at EC reg 0x30,
> fw 7XHT24WW-1.06'
> Mixer name : 'ThinkPad EC 7XHT24WW-1.06'
> Components : ''
> Controls : 1
> Simple ctrls : 1
> Card29.Amixer.values:
> Simple mixer control 'Console',0
> Capabilities: pswitch pswitch-joined penum
> Playback channels: Mono
> Mono: Playback [on]
> Date: Wed Apr 13 15:03:35 2011
> EcryptfsInUse: Yes
> HibernationDevice: RESUME=UUID=bb479652-f524-4abe-b1b0-27646d6deebc
> InstallationMedia: Ubuntu 10.10 "Maverick Meerkat" - Release Candidate
> amd64 (20100928)
> MachineType: LENOVO 7465CTO
> ProcEnviron:
> LANGUAGE=en_US:en_GB:en
> PATH=(custom, user)
> LANG=en_US.UTF-8
> SHELL=/bin/bash
> ProcKernelCmdLine: BOOT_IMAGE=/boot/vmlinuz-2.6.38-8-generic
> root=UUID=266abe9a-19c2-4cc3-9ef7-238b729b6044 ro quiet splash
> libata.force=noncq vt.handoff=7
> RelatedPackageVersions:
> linux-restricted-modules-2.6.38...

[Newfie (mn-newfie)]

I should chime in here to mention that battery life for me on Natty was very negatively impacted. The most I was able to get out of my laptop when dimming screen, disabling effects, etc, was 50 minutes.

The same laptop can get a max of 2.5 hours on Ubuntu 10.10 and lower.

[Nicolas Michel (nicolas-michel)]

On 06/28/2011 09:49 AM, Newfie wrote:
> I should chime in here to mention that battery life for me on Natty was
> very negatively impacted. The most I was able to get out of my laptop
> when dimming screen, disabling effects, etc, was 50 minutes.
>
> The same laptop can get a max of 2.5 hours on Ubuntu 10.10 and lower.
>
Have you tried the fix proposed by Michael?
-> add to your kernel option (into grub) pcie_aspm=force

If you don't know how to do it, you'll find instructions into previous
comments of this bug report.

[Peter Sasi (peter-sasi)]

I have added pcie_aspm=force to my kernel (2.6.38, Natty) command line and rebooted. It is picked up:
"PCIe ASPM is forcedly enabled"
It still seems diabled:
"ACPI FADT declares the system doesn't support PCIe ASPM, so disable it"
lspci shows its only enabled for the controller and the nvidia card:
:~$ sudo lspci -vvv|grep -i aspm
  LnkCap: Port #2, Speed 2.5GT/s, Width x16, ASPM L0s L1, Latency L0 <256ns, L1 <4us
  LnkCtl: ASPM L0s L1 Enabled; RCB 64 bytes Disabled- Retrain- CommClk+
  LnkCap: Port #0, Speed unknown, Width x0, ASPM unknown, Latency L0 <64ns, L1 <1us
  LnkCtl: ASPM Disabled; Disabled- Retrain- CommClk-
  LnkCap: Port #1, Speed 2.5GT/s, Width x1, ASPM L0s L1, Latency L0 <1us, L1 <4us
  LnkCtl: ASPM Disabled; RCB 64 bytes Disabled- Retrain- CommClk-
  LnkCap: Port #2, Speed 2.5GT/s, Width x1, ASPM L0s L1, Latency L0 <256ns, L1 <4us
  LnkCtl: ASPM Disabled; RCB 64 bytes Disabled- Retrain- CommClk+
  LnkCap: Port #3, Speed 2.5GT/s, Width x1, ASPM L0s L1, Latency L0 <1us, L1 <4us
  LnkCtl: ASPM Disabled; RCB 64 bytes Disabled- Retrain- CommClk-
  LnkCap: Port #4, Speed 2.5GT/s, Width x1, ASPM L0s L1, Latency L0 <1us, L1 <4us
  LnkCtl: ASPM Disabled; RCB 64 bytes Disabled- Retrain- CommClk-
  LnkCap: Port #5, Speed 2.5GT/s, Width x1, ASPM L0s L1, Latency L0 <1us, L1 <4us
  LnkCtl: ASPM Disabled; RCB 64 bytes Disabled- Retrain- CommClk-
  LnkCap: Port #0, Speed 2.5GT/s, Width x16, ASPM L0s L1, Latency L0 <512ns, L1 <4us
  LnkCtl: ASPM L0s L1 Enabled; RCB 128 bytes Disabled- Retrain- CommClk+
  LnkCap: Port #0, Speed 2.5GT/s, Width x1, ASPM L0s L1, Latency L0 <128ns, L1 <64us
  LnkCtl: ASPM Disabled; RCB 64 bytes Disabled- Retrain- CommClk+

[Olcay Korkmaz (olci)]

Acer aspire 5315 withnatty on 2.6.38 pcie_aspm=force enabled on dmesg shows "PCIe ASPM is forcedly enabled" but

only wireless device'aspm has been enabled
"broadcom bcm4311bg rev1 "

others still disabled

LnkCap: Port #0, Speed unknown, Width x0, ASPM unknown, Latency L0 <64ns, L1 <1us
LnkCtl: ASPM Disabled; Disabled- Retrain- CommClk-
LnkCap: Port #1, Speed 2.5GT/s, Width x1, ASPM L0s L1, Latency L0 <1us, L1 <4us
LnkCtl: ASPM Disabled; RCB 64 bytes Disabled- Retrain- CommClk-
LnkCap: Port #2, Speed 2.5GT/s, Width x1, ASPM L0s L1, Latency L0 <1us, L1 <4us
LnkCtl: ASPM Disabled; RCB 64 bytes Disabled- Retrain- CommClk-
LnkCap: Port #3, Speed 2.5GT/s, Width x1, ASPM L0s L1, Latency L0 <256ns, L1 <4us
LnkCtl: ASPM Disabled; RCB 64 bytes Disabled- Retrain- CommClk+
LnkCap: Port #4, Speed 2.5GT/s, Width x1, ASPM L0s L1, Latency L0 <256ns, L1 <4us
LnkCtl: ASPM Disabled; RCB 64 bytes Disabled- Retrain- CommClk+
LnkCap: Port #0, Speed 2.5GT/s, Width x1, ASPM L0s L1, Latency L0 <4us, L1 <64us
LnkCtl: ASPM Disabled; RCB 64 bytes Disabled- Retrain- CommClk+
LnkCap: Port #0, Speed 2.5GT/s, Width x1, ASPM L0s, Latency L0 <4us, L1 <64us
LnkCtl: ASPM L0s Enabled; RCB 64 bytes Disabled- Retrain- CommClk+

[Sam_ (and-sam)]

PCIe ASPM is forcedly enabled, but battery decreases rapidly, disabled it again. (HP 625)

[srd (sr-gimp)]

A slight decrease in power usage, dropping from 33 to 30.5 Watts according to powertop on a Thinkpad R500. Better than nothing, but still a long way from the sub 20 Watts I used to be able to get with ubuntu 10.04 and before.

[Marko Martinović (marko-martinovic-deactivatedaccount)]

Hi guys, I've had my HP Compaq CQ56 freezing when I've enabled pcie_aspm=force so you can test using sysfs to check would enabling it make problems before enabling pcie_aspm=force for good like I've wrote here:

http://www.techytalk.info/2011/06/linux-kernel-2-6-38-2-6-39-power-regression-workaround/

My findings on my old ASUS laptop are that there is a slight increase in battery life but not near to Lucid and its 2.6.32 kernel :(

[Matthew Hessel (matt-hessel)]

Are you using the stock kernel? Or the patched version further up the thread?

On 6/29/11, srd <email address hidden> wrote:
> A slight decrease in power usage, dropping from 33 to 30.5 Watts
> according to powertop on a Thinkpad R500. Better than nothing, but still
> a long way from the sub 20 Watts I used to be able to get with ubuntu
> 10.04 and before.
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/760131
>
> Title:
> Power consumption raised significantly in natty
>
> Status in Release Notes for Ubuntu:
> Fix Released
> Status in “linux” package in Ubuntu:
> Triaged
> Status in “linux” source package in Natty:
> Incomplete
> Status in “linux” source package in Oneiric:
> Triaged
>
> Bug description:
> This is a regression moving from maverick->natty. With the same sort
> of light interactive usage, Battery life has gone from ~4 hours to
> 2:45 or so.
>
> Running powertop shows the system in P-state "Turbo mode" 10-15% of
> the time. Typically this was almost zero in maverick, 98+% of the
> time maverick was in the lowest state.
>
> wakeups reported in the 5-600/s range where previously (from memory)
> it was 200 ish. i915 driver is high on the list. This is
> irrespective of whether running unity or classic desktop.
>
> ProblemType: Bug
> DistroRelease: Ubuntu 11.04
> Package: linux-image-2.6.38-8-generic 2.6.38-8.42
> Regression: Yes
> Reproducible: Yes
> ProcVersionSignature: Ubuntu 2.6.38-8.42-generic 2.6.38.2
> Uname: Linux 2.6.38-8-generic x86_64
> AlsaVersion: Advanced Linux Sound Architecture Driver Version 1.0.23.
> Architecture: amd64
> ArecordDevices:
>  **** List of CAPTURE Hardware Devices ****
>  card 0: Intel [HDA Intel], device 0: CONEXANT Analog [CONEXANT Analog]
>    Subdevices: 1/1
>    Subdevice #0: subdevice #0
> AudioDevicesInUse:
>  USER PID ACCESS COMMAND
>  /dev/snd/controlC0: james 2068 F.... pulseaudio
> CRDA: Error: [Errno 2] No such file or directory
> Card0.Amixer.info:
>  Card hw:0 'Intel'/'HDA Intel at 0xf2620000 irq 47'
>    Mixer name : 'Conexant CX20561 (Hermosa)'
>    Components : 'HDA:14f15051,17aa20ff,00100000'
>    Controls : 16
>    Simple ctrls : 8
> Card29.Amixer.info:
>  Card hw:29 'ThinkPadEC'/'ThinkPad Console Audio Control at EC reg 0x30,
> fw 7XHT24WW-1.06'
>    Mixer name : 'ThinkPad EC 7XHT24WW-1.06'
>    Components : ''
>    Controls : 1
>    Simple ctrls : 1
> Card29.Amixer.values:
>  Simple mixer control 'Console',0
>    Capabilities: pswitch pswitch-joined penum
>    Playback channels: Mono
>    Mono: Playback [on]
> Date: Wed Apr 13 15:03:35 2011
> EcryptfsInUse: Yes
> HibernationDevice: RESUME=UUID=bb479652-f524-4abe-b1b0-27646d6deebc
> InstallationMedia: Ubuntu 10.10 "Maverick Meerkat" - Release Candidate
> amd64 (20100928)
> MachineType: LENOVO 7465CTO
> ProcEnviron:
>  LANGUAGE=en_US:en_GB:en
>  PATH=(custom, user)
>  LANG=en_US.UTF-8
>  SHELL=/bin/bash
> ProcKernelCmdLine: BOOT_IMAGE=/boot/vmlinuz-2.6.38-8-generic
> root=UUID=266abe9a-19c2-4cc3-9ef7-2...

[Sam_ (and-sam)]

Forgot to add on #116, running Oneiric with 3.0-2 kernel. Not a techie here, hence unsure if output is ok (without pcie_aspm=force).

[lljccoael (lljccoael-deactivatedaccount)]

Adding 'pcie_aspm=force' to stock Natty install on a Toshiba L305-S5937 decreased the heat significantly. This laptop is now usable in Natty without literally burning. Battery consumption has also increased from one to about two hours. I have observed no freezing after the latest stable updates. I have not yet tested hard 3d performance, like a game or something.

However, the 2.6.35 regression remains, and it looks like Michael Larabel is the only one telling users what's going on, so we'll have to wait for him.

Even though it is claimed by some this is not a regression, but a feature, obviously a lot of portable computers were devastated by it, so it's absolutely a regression.

@felixcorrales: You have not been told everything.

You can 'pcie_aspm=force' to

ALT-F2 > 'gksu nautilus', press enter > open File System /etc/default/grub file >

GRUB_CMDLINE_LINUX_DEFAULT=”quiet splash”
GRUB_CMDLINE_LINUX=”"

You can add 'pcie_aspm=force' to GRUB_CMDLINE_LINUX_DEFAULT after "splash" within the quotes or GRUB_CMDLINE_LINUX within the quotes. Save, then *******open Terminal and type 'sudo update-grub' without quotes, press enter, wait for it to finish, then reboot the computer. Only then will the command apply to the kernel.*******

[carlosv (cvedovatti)]

Hi there,
I certainly agree with cjcolella. It is a clear regression. I have a HP touchsmart tm2, intel i5 core. With Ubuntu 10.04 I had between 5 and 6 hours of battery. With Ubuntu 11.04 (Kernel 2.6.38-8-generic), between 2,5 and 3 hours. It is a really huge problem for this distribution.
I just tried adding "pcie_aspm=force". In my case there is no change with the power consumption but I notice that it helped to decrease the temperature on the computer. Now my lap is not burning when I use it, but battery life still the same.

[Jon Grossart (jon-grossart)]

While, I agree that this a huge problem overall, I'm think I have to agree with the idea that it's not a regression. If you read the comments on Phoronix from the actual path author (which was only linked on the little side bar of the page, not the main column), he makes the statement a little more clear.

http://www.phoronix.com/scan.php?page=news_item&px=OTYwNA

The patch enables correct handling of the ASPM to the BIOS if it can handle it, otherwise it lets the kernel do it. However, it then depends on the hardware drivers to setup the correct values. That is the problem that is happening with this bug according to the author. The kernel is now handling the situation correctly, and it is revealing a lot of long standing bugs in a lot/most of the drivers and BIOSes in that they don't handle ASPM correctly, or coded to the fact that the kernel was incorrectly managing it before. Quite frankly, it sounds a lot like the whole IE6 problem plaguing web browsers.

Maybe this has something to do with the overall higher power consumption in Ubuntu often seen. On netbook in WinXP I can it down to 5W idling. In Ubuntu with same config, it won't go below 7-8W. The only thing I can think that would make the different is the driver settings/abilities.

So what is the correct fix? That, I don't know. BIOS and hardware drivers are real culprits it seems. Chances of BIOS fixes are probably slim for the vast majority of systems. And the drivers changes would probably be numerous and slow going.

[Matthew Hessel (matt-hessel)]

I would offer the opinion that this is more like a regression as this
exposed a kernel regression as proper aspm handling requires changes
to the drivers in the kernel.

In the article, it mentions that the windows drivers may have
additional bits for aspm functionality, thus overriding BIOS handover.

This is not unlike previous issues seen with linux where many BIOS
implementations are usually just good enough to get windows working,
specific functions they are supposed to do are not done according to
standards.

Before this issue there were numerous issues with suspend/resume,
DPMI, intel graphics bios missing resolution info, and more. Each
issue resulted in hacks to work around until drivers or the kernel got
a more permanent fix.

On 6/30/11, Jon Grossart <email address hidden> wrote:
> While, I agree that this a huge problem overall, I'm think I have to
> agree with the idea that it's not a regression. If you read the
> comments on Phoronix from the actual path author (which was only linked
> on the little side bar of the page, not the main column), he makes the
> statement a little more clear.
>
> http://www.phoronix.com/scan.php?page=news_item&px=OTYwNA
>
> The patch enables correct handling of the ASPM to the BIOS if it can
> handle it, otherwise it lets the kernel do it. However, it then depends
> on the hardware drivers to setup the correct values. That is the
> problem that is happening with this bug according to the author. The
> kernel is now handling the situation correctly, and it is revealing a
> lot of long standing bugs in a lot/most of the drivers and BIOSes in
> that they don't handle ASPM correctly, or coded to the fact that the
> kernel was incorrectly managing it before. Quite frankly, it sounds a
> lot like the whole IE6 problem plaguing web browsers.
>
> Maybe this has something to do with the overall higher power consumption
> in Ubuntu often seen. On netbook in WinXP I can it down to 5W idling.
> In Ubuntu with same config, it won't go below 7-8W. The only thing I
> can think that would make the different is the driver
> settings/abilities.
>
> So what is the correct fix? That, I don't know. BIOS and hardware
> drivers are real culprits it seems. Chances of BIOS fixes are probably
> slim for the vast majority of systems. And the drivers changes would
> probably be numerous and slow going.
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/760131
>
> Title:
> Power consumption raised significantly in natty
>
> Status in Release Notes for Ubuntu:
> Fix Released
> Status in “linux” package in Ubuntu:
> Triaged
> Status in “linux” source package in Natty:
> Incomplete
> Status in “linux” source package in Oneiric:
> Triaged
>
> Bug description:
> This is a regression moving from maverick->natty. With the same sort
> of light interactive usage, Battery life has gone from ~4 hours to
> 2:45 or so.
>
> Running powertop shows the system in P-state "Turbo mode" 10-15% of
> the time. Typically this was almost zero in maverick, 98+% of the
> time maverick was in the lowest state.
>
> wakeups reported in the...

[Jon Grossart (jon-grossart)]

Regression/proper behavior or whatever it ends up being called, there is a lot more fixing that needs to be done. What has surprised me about the whole Linux community is that no one seems to be really all that upset or even tracking the problem as a whole except Phoronix and the users. It would seem with the prevalence of laptops that power consumption would be a huge priority.

Although, there are tons more servers than laptop running Linux, so maybe that is the issue here as well. Although, I would think power consumption is an even bigger issue for them (cost wise, not running down the battery/burn your lap wise).

[Matthew Hessel (matt-hessel)]

Jon- absolutely agree. Only way I can figure this isn't getting more
attention is that the majority of datacenters are using enterprise
versions of the kernel, which are several versions back and aren't
affected yet.

Once RHEL version 7 or 8 comes out with 2.6.38 or later, everyone
would get a bit concerned when the power utilization goes up.

FWIW it probably wouldn't affect really busy servers as much, this
seems to really affect power usage when the cpu is lightly loaded or
idle.

On 6/30/11, Jon Grossart <email address hidden> wrote:
> Regression/proper behavior or whatever it ends up being called, there is
> a lot more fixing that needs to be done. What has surprised me about
> the whole Linux community is that no one seems to be really all that
> upset or even tracking the problem as a whole except Phoronix and the
> users. It would seem with the prevalence of laptops that power
> consumption would be a huge priority.
>
> Although, there are tons more servers than laptop running Linux, so
> maybe that is the issue here as well. Although, I would think power
> consumption is an even bigger issue for them (cost wise, not running
> down the battery/burn your lap wise).
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/760131
>
> Title:
> Power consumption raised significantly in natty
>
> Status in Release Notes for Ubuntu:
> Fix Released
> Status in “linux” package in Ubuntu:
> Triaged
> Status in “linux” source package in Natty:
> Incomplete
> Status in “linux” source package in Oneiric:
> Triaged
>
> Bug description:
> This is a regression moving from maverick->natty. With the same sort
> of light interactive usage, Battery life has gone from ~4 hours to
> 2:45 or so.
>
> Running powertop shows the system in P-state "Turbo mode" 10-15% of
> the time. Typically this was almost zero in maverick, 98+% of the
> time maverick was in the lowest state.
>
> wakeups reported in the 5-600/s range where previously (from memory)
> it was 200 ish. i915 driver is high on the list. This is
> irrespective of whether running unity or classic desktop.
>
> ProblemType: Bug
> DistroRelease: Ubuntu 11.04
> Package: linux-image-2.6.38-8-generic 2.6.38-8.42
> Regression: Yes
> Reproducible: Yes
> ProcVersionSignature: Ubuntu 2.6.38-8.42-generic 2.6.38.2
> Uname: Linux 2.6.38-8-generic x86_64
> AlsaVersion: Advanced Linux Sound Architecture Driver Version 1.0.23.
> Architecture: amd64
> ArecordDevices:
>  **** List of CAPTURE Hardware Devices ****
>  card 0: Intel [HDA Intel], device 0: CONEXANT Analog [CONEXANT Analog]
>    Subdevices: 1/1
>    Subdevice #0: subdevice #0
> AudioDevicesInUse:
>  USER PID ACCESS COMMAND
>  /dev/snd/controlC0: james 2068 F.... pulseaudio
> CRDA: Error: [Errno 2] No such file or directory
> Card0.Amixer.info:
>  Card hw:0 'Intel'/'HDA Intel at 0xf2620000 irq 47'
>    Mixer name : 'Conexant CX20561 (Hermosa)'
>    Components : 'HDA:14f15051,17aa20ff,00100000'
>    Controls : 16
>    Simple ctrls : 8
> Card29.Amixer.info:
>  Card hw:29 'ThinkPad...

[florin (florin-arjocu)]

As cjcolella said, for people that switched to Ubuntu recently, just like me, the line in /etc/default/grub looks like this now (Natty, all updates):

GRUB_CMDLINE_LINUX_DEFAULT="quiet splash pcie_aspm=force"

My Toshiba laptop (C650-16R, Intel i3, intel hd graphics) can now run longer, but I agree with all people saying this is a huge bug. On first run of Ubuntu I could enjoy the laptop almost 4 hours and a few days ago it only worked for more that one hour. Now I get almost 2 hours (better, but not perfect). I hope the battery is not permanently affected.

[RussianNeuroMancer (russianneuromancer)]

http://lwn.net/Articles/449648/

[Warwick Bruce Chapman (warwickchapman)]

I'm on a Lenovo X301. dmesg says this system does not support ASPM but battery life is still poor.

-- 083 7797 094 | http://warwickchapman.com

-----Original Message-----
From: RussianNeuroMancer <email address hidden>
Sender: <email address hidden>
Date: Fri, 01 Jul 2011 02:33:56
To: <email address hidden>
Reply-To: Bug 760131 <email address hidden>
Subject: [Bug 760131] Re: Power consumption raised significantly in natty

http://lwn.net/Articles/449648/

--
You received this bug notification because you are subscribed to the bug
report.
https://bugs.launchpad.net/bugs/760131

Title:
  Power consumption raised significantly in natty

Status in Release Notes for Ubuntu:
  Fix Released
Status in “linux” package in Ubuntu:
  Triaged
Status in “linux” source package in Natty:
  Incomplete
Status in “linux” source package in Oneiric:
  Triaged

Bug description:
  This is a regression moving from maverick->natty. With the same sort
  of light interactive usage, Battery life has gone from ~4 hours to
  2:45 or so.

  Running powertop shows the system in P-state "Turbo mode" 10-15% of
  the time. Typically this was almost zero in maverick, 98+% of the
  time maverick was in the lowest state.

  wakeups reported in the 5-600/s range where previously (from memory)
  it was 200 ish. i915 driver is high on the list. This is
  irrespective of whether running unity or classic desktop.

  ProblemType: Bug
  DistroRelease: Ubuntu 11.04
  Package: linux-image-2.6.38-8-generic 2.6.38-8.42
  Regression: Yes
  Reproducible: Yes
  ProcVersionSignature: Ubuntu 2.6.38-8.42-generic 2.6.38.2
  Uname: Linux 2.6.38-8-generic x86_64
  AlsaVersion: Advanced Linux Sound Architecture Driver Version 1.0.23.
  Architecture: amd64
  ArecordDevices:
   **** List of CAPTURE Hardware Devices ****
   card 0: Intel [HDA Intel], device 0: CONEXANT Analog [CONEXANT Analog]
     Subdevices: 1/1
     Subdevice #0: subdevice #0
  AudioDevicesInUse:
   USER PID ACCESS COMMAND
   /dev/snd/controlC0: james 2068 F.... pulseaudio
  CRDA: Error: [Errno 2] No such file or directory
  Card0.Amixer.info:
   Card hw:0 'Intel'/'HDA Intel at 0xf2620000 irq 47'
     Mixer name : 'Conexant CX20561 (Hermosa)'
     Components : 'HDA:14f15051,17aa20ff,00100000'
     Controls : 16
     Simple ctrls : 8
  Card29.Amixer.info:
   Card hw:29 'ThinkPadEC'/'ThinkPad Console Audio Control at EC reg 0x30, fw 7XHT24WW-1.06'
     Mixer name : 'ThinkPad EC 7XHT24WW-1.06'
     Components : ''
     Controls : 1
     Simple ctrls : 1
  Card29.Amixer.values:
   Simple mixer control 'Console',0
     Capabilities: pswitch pswitch-joined penum
     Playback channels: Mono
     Mono: Playback [on]
  Date: Wed Apr 13 15:03:35 2011
  EcryptfsInUse: Yes
  HibernationDevice: RESUME=UUID=bb479652-f524-4abe-b1b0-27646d6deebc
  InstallationMedia: Ubuntu 10.10 "Maverick Meerkat" - Release Candidate amd64 (20100928)
  MachineType: LENOVO 7465CTO
  ProcEnviron:
   LANGUAGE=en_US:en_GB:en
   PATH=(custom, user)
   LANG=en_US.UTF-8
   SHELL=/bin/bash
  ProcKernelCmdLine: BOOT_IMAGE=/boot/vmlinuz-2.6.38-8-ge...

[michaellarabel (michael-michaellarabel)]

New Effort: http://www.phoronix.com/scan.php?page=news_item&px=OTYxOQ

[florin (florin-arjocu)]

Don't throw your stones for this idea, started by this post http://lwn.net/Articles/449648/ :D

Can Linux test the power usage of the computer/device during install or have periodic tests? This looks like a trial and error test, but I think it might just work.

Something like:
no pcie_aspm
test power usage/wake ups {different scenarios}
activate pcie_aspm
test power usage/wake ups {same different scenarios}
chose best solution and remember settings on updates, send data to centralized database if user allows it.

[RussianNeuroMancer (russianneuromancer)]

It's indeed intresting idea. Distribution with large user-base like Ubuntu really can done that.

[felixcorrales (felixcorrales-yahoo)]

I have tested 'pcie_aspm=force'

Lenovo T400s
Intel Core 2 Duo CPU P9600
Integrated Intel 4500MHD GPU

>sudo powertop

With 'pcie_aspm=force'

Consumo de energía (estimado por ACPI): 14,8W (2,3 horas) (largo plazo: 14,6W,/2,3h) 4:21 pm 48/47 C
Consumo de energía (estimado por ACPI): 14,7W (2,3 horas) (largo plazo: 14,8W,/2,2h) 4:24 pm 45/41 C
Consumo de energía (estimado por ACPI): 14,7W (2,2 horas) (largo plazo: 14,6W,/2,2h) 4:28 pm 48/44 C
Consumo de energía (estimado por ACPI): 14,8W (2,2 horas) (largo plazo: 14,7W,/2,2h) 4:40 pm 46/42 C
Consumo de energía (estimado por ACPI): 14,4W (2,1 horas) (largo plazo: 14,7W,/2,1h) 4:33 pm 44/40 C

Without 'pcie_aspm=force'

Consumo de energía (estimado por ACPI): 15,8W (2,1 horas) (largo plazo: 15,9W,/2,1h) 4:15 pm 45/41 C
Consumo de energía (estimado por ACPI): 15,4W (2,1 horas) (largo plazo: 15,8W,/2,1h) 4:17 pm 43/39 C
Consumo de energía (estimado por ACPI): 15,2W (2,1 horas) (largo plazo: 15,8W,/2,0h) 4:18 pm 43/39 C
Consumo de energía (estimado por ACPI): 15,3W (2,1 horas) (largo plazo: 15,3W,/2,0h) 4:21 pm 48/44 C
Consumo de energía (estimado por ACPI): 15,9W (2,0 horas) (largo plazo: 15,5W,/2,0h) 4:22 pm 45/41 C
Consumo de energía (estimado por ACPI): 15,3W (2,0 horas) (largo plazo: 15,5W,/2,0h) 4:24 pm 44/40 C
Consumo de energía (estimado por ACPI): 15,4W (2,0 horas) (largo plazo: 15,4W,/2,0h) 4:25 pm 43/40 C

Conclusion:

The power consumption has been decreased by 1 Watt/h with 'pcie_aspm=force'

[SK Sullivan (78luphr0rnk-launchpad-a811i2i3ytq)]

Dell Latitude laptop L13, Intel Core 2 Duo SU7300, Ubuntu 11.04

After going to Natty, battery life tanked (not that it was so great to begin with, though).

I picked up about an extra hour of battery life after adding pcie_aspm=force to grub. Was getting about 2.5 hours, now get just under 3.5 hours.

Could just be n-rays, lol. Also seems much cooler... :)

[Adrian Wechner (adrian-wechner)]

adding pcie_aspm=force

ubuntu 11.04
kernel 2.6.38-8-generic
Dell inspiron 1525
from 19.0W to 15.0W down (idle). 20% less.

[carlosv (cvedovatti)]

I just tried all way possible my HP touchsmart tm2 in idle.

Ubuntu 10.10 I got 5 hours of battery.

Ubuntu 11.04 without pcie_aspm=force I got 3 hours of battery.

Ubuntu 11.04 with pcie_aspm=force I got 4,5 hours of battery.

With windows 7 I got 6,5 hours of battery.

It's a really big problem for this distribution I just hope that will be solved for Oneric!

[lljccoael (lljccoael-deactivatedaccount)]

There won't be a fix per se for Oneiric. This is a permanent condition because it is a new setting of the Linux kernel, and nobody in power considers it a bug, nor a very important issue. Sorry.

"There isn't any easy 'solution' to improve this situation beyond affected users forcing the PCI-E Active-State Power Management using the pcie_aspm=force kernel command line option. Jesse thinks that more Linux drivers will end up needing to set the ASPM bits directly as a long-term solution. There's just too many hardware devices that don't properly support the ASPM power-saving modes.

The only alternative would be to create a big white-list of supported devices, but that comes down to being effectively the same large task as just having the driver set the appropriate bit. So there won't be any magic fix in the Linux 3.0 kernel nor will there likely be any major change in the Linux 3.1 kernel without suddenly a bunch of drivers handling the Active-State Power Management bit. For now, mobile users just need to know to force the PCI-E ASPM support if needed to maximize the Linux battery life."

[Nicolas Michel (nicolas-michel)]

Out of curiosity:

Why Linux have this problem and Windows don't?
This is not a troll. I don't understand technically : if hardware
devices don't propertly support the ASPM power-saving modes, why Windows
is not concerned?

On 07/18/2011 02:24 AM, cjcolella wrote:
> There won't be a fix per se for Oneiric. This is a permanent condition
> because it is a new setting of the Linux kernel, and nobody in power
> considers it a bug, nor a very important issue. Sorry.
>
> "There isn't any easy 'solution' to improve this situation beyond
> affected users forcing the PCI-E Active-State Power Management using the
> pcie_aspm=force kernel command line option. Jesse thinks that more Linux
> drivers will end up needing to set the ASPM bits directly as a long-term
> solution. There's just too many hardware devices that don't properly
> support the ASPM power-saving modes.
>
> The only alternative would be to create a big white-list of supported
> devices, but that comes down to being effectively the same large task as
> just having the driver set the appropriate bit. So there won't be any
> magic fix in the Linux 3.0 kernel nor will there likely be any major
> change in the Linux 3.1 kernel without suddenly a bunch of drivers
> handling the Active-State Power Management bit. For now, mobile users
> just need to know to force the PCI-E ASPM support if needed to maximize
> the Linux battery life."
>

[Derek P (derekpop)]

Novice user adding input here. If this cannot be fixed so that I can have similar battery life as when running Windows I will have to stop using Ubuntu. I prefer this OS over Windows 7 but I can't work with barely 2 hours of battery life. This is with - GRUB_CMDLINE_LINUX_DEFAULT="quiet splash pcie_aspm=force"

I am crossing my fingers they figure something out. Thanks for listening.

[florin (florin-arjocu)]

I base my opinion on +cjcolella's note:

I do think that not considering battery life as a priority will make Linux unusable in the near future. Why? Just have a look at Ubuntu's interface and one of Ubuntu's major decisions: it dropped Gnome and made a whole new interface, Unity, that has one of the main objectives to make it EASY TO USE on touchscreens.

But most touch screens are on phones, tablets, notebooks, netbooks, and other mobile devices. I think nowadays, mobile devices are the most succesful on any markets and I think Ubuntu had a good idea to address the mobile users; but it's important to make the OS work at least as well as other OS's.

Why would I use Ubuntu/any other Linux (as it is a kernel "feature" aka bug) if my cool device can't be used with it?

Call me negativist, but this is a any-linux-distro-killer issue and should be shared with other distributions and insist at kernel developers. This needs a solution ASAP, before users start looking elsewhere.

[Matthew Hessel (matt-hessel)]

that is the interesting question.

There isn't really a large discrepancy for the most part.

Phoronix did a piece comparing ubuntu to windows 7, and they actually were
rather close.

http://www.phoronix.com/scan.php?page=article&item=windows_ubuntu_pow&num=1

that said, there are a few reasons why Natty ends up using more power than
previous versions..

the biggest is the ASPI changes to the kernel. Basically, there are some
devices that could lock up the boot process as earlier versions would boot
up. The kernel devs changed the behavior of the linux kernel, now it does
not assume it can implement ASPM, it will query the ACPI bits in the BIOS on
your computer if it can. Before this, it didn't check, it would turn it on
regardless.

Unfortunately, as these things go, many BIOS out there SUCK. Most
manufacturers do little testing (as in none) to make sure their bios works
according to established standards, and usually if it works with Windows,
they ship it, and don't really care afterwards. Now that the Linux kernel
is being polite, (and following actual ACPI standards) many devices will
not report proper information to hand off to the Kernel. So Linux will now
assume it can't touch it, and ASPM is disabled on the devices.

for the most part, you can override this with the pcie_aspm=force command
during boot. There does seem to be some devices I have that don't enable
ASPM even with that for me, mostly some usb stuff on my laptop and the Intel
wireless 5100 adapter. I don't know that it makes a huge difference, with
the Maverick kernel, I can see idle power usage down to 18 watts, but that
is all PM options enabled, and nothing running but Unity (by the way, unity
is much more power efficient than the older Gnome 2.x - Who would have
thought..)

with the latest upstream sources from git.kernel.org or the Oneiric sources
I can see it idle at about 19-20 watts. Not a lot lost.

the other thing to realize is there were many scheduler changes that have
been made since 2.6.38 in Natty. Many of them improve things for power
utilization. One developer found that there was a problem with counters
rolling over that would cause timing issues with load balancing between
multiple cores, and it would wake the CPUs early, preventing them from
idling in low power states.

This is fixed in the upstream sources, and helps a bit with the overall
performance. Maybe it is just me, but when I am actually doing something on
the laptop, it doesn't seem to run the battery down as bad with the patch
in, so maybe that is something.

Your experience may vary, some didn't report any issues in this thread at
all with natty, some even said they got BETTER life out of the battery on
Natty vs others.

I think it depends more on how seriously your laptop manufacturer takes BIOS
debugging, as many don't try very hard. If you happen to be a lucky person
with a good BIOS, like maybe a decent lenovo or something, then you might
not see any issues at all.

Matthew Garrett was the subject of a recent article on LWN.net that gives a
good history on this. Maybe it will help more than my rambling.

http://lwn.net/Articles/449448/

Most of the stuff I have read from th...

[Matthew Hessel (matt-hessel)]

well, I woudn't say negativist, but..

I think the power issue is a bit overblown. even in the inital bug report
for this, Natty was like 15 percent more power hungry than the older
versions. And - this is not unique to Natty, or Ubuntu..

if you look around there are bug reports for this exact bug on Fedora,
Debian, I think I even saw this in Gentoo and Arch as well. Most of the
effort I see Ubuntu developers working on are usability and GUI
improvements, and don't discount the work they do to enhance the stock
debian packages to make Ubuntu as stable as possible.

the Power utilization is higher, true, newer kernels do not have as much of
a problem, and there is the option to force the pcie_aspm to work like it
used to, so there is no functionality lost. It is a bit of a cludge, but
you should expect that when running the latest and more bleeding edge stuff.

Before you start assuming that there is a major disconnect with the guys at
ubuntu though, try out the latest Fedora.. notice anything?

Power issues are the same, and they have this fancy new GUI also, only it is
Gnome 3.0 - I'm not saying either one is better, but there are more
similarities than differences that I have found. But lately Gnome has
ticked me off too much - you have LESS options to customize and tune than
Unity or others. I used fedora for maybe a week, reinstalled Natty and have
been happy since.

If you don't like either one, and want something more flexible or whatever,
try the alternatives. KDE is looking really good these days, might be the
prettiest GUI I have ever seen, and now it really seems usable, how funny
that a year ago, everyone was preaching doom and gloom on the changes from
KDE old to KDE new.

I will also say that Ubuntu is one of the more stable and easy to use
distros out there, they don't take away my insatiable desire to tinker with
the guts of it. And I don't see any other OS that is as easy to add other
devices to now. Windows is a pain in the @#$@ to get some new device on
there, what with the installers that HP wants you to use so you can't just
install the damn thing, - I have to also add your bloated replacement for
stuff that windows can already do, and other issues..

Macs might "just work," but only if you use their stuff.. A PC would just
work too if I only had maybe 20 or so things I could plug into it..

anyway I have veered way off the subject.. hope I don't sound too
trollish..

On Mon, Jul 18, 2011 at 7:01 PM, florin <email address hidden> wrote:

> I base my opinion on +cjcolella's note:
>
> I do think that not considering battery life as a priority will make
> Linux unusable in the near future. Why? Just have a look at Ubuntu's
> interface and one of Ubuntu's major decisions: it dropped Gnome and made
> a whole new interface, Unity, that has one of the main objectives to
> make it EASY TO USE on touchscreens.
>
>
> But most touch screens are on phones, tablets, notebooks, netbooks, and
> other mobile devices. I think nowadays, mobile devices are the most
> succesful on any markets and I think Ubuntu had a good idea to address the
> mobile users; but it's important to make the OS work at least as well as
> other OS...

[florin (florin-arjocu)]

+Matthew Hessel

"if you look around there are bug reports for this exact bug on Fedora,
Debian, I think I even saw this in Gentoo and Arch as well. Most of the
effort I see Ubuntu developers working on are usability and GUI
improvements, and don't discount the work they do to enhance the stock
debian packages to make Ubuntu as stable as possible."

That's what I said, it's a kernel bug, so it should be in all linux distributions. Also that's the reason I suggested Ubuntu developers should work together with all other distributions' developers, to force a faster and better solution.
I think it's one of the times when it does not matter what color we dress up, but it matters for the same kernel we all use to work better :)

I'm using a Toshiba laptop and installed Ubuntu 10.10 in March (this year), then upgraded to 11.4, then reinstalled 11.4 as at least for me it worked better as a fresh install (maybe newbie issues) :D
The thing is that the first installation made me enjoy a lot more time without plugging the machine, and during later updates I could only use it less than one hour (1h). I guess you can image it's frustrating.
The pcie_aspm=force gave me like a plus of about 30 minutes, so it's definitely a good idea.

This problem might not be the most important in the world, but as much as I can think, it's a huge problem for all mobile devices and a regular user does not know how or should not have to insert some manual parameters to grub. Imagine my grandmother who has first seen a keyboard 2 years ago doing that; but of course she can plug or unplug the laptop to/from electricity line as that is like an old radio, you don't need to know anything complicated for that.

In the end I just want to make sure people know I appreciate what Ubuntu team is doing. The're doing a great work, I could see things being fixed in the last months (for instance until later Natty with updates, the headphones were not working, but they fixed it; don't know if it's only on a virtual machine, but there is another bug fixed in Oneiric, alpha: disabling network hangs system in Natty, but not in Oneric). I do what I can to help, for instance I've reported some bugs and I can definitely do more testing as I know what to expect from a machine and I use the laptop 15+ hours a day.

[Sam (smickson)]

@ Matthew Hessel

If I buy a Lenovo Thinkpad netbook/laptop, am I pretty much certain to get a "good" BIOS? I'm going to be buying a new computer. How can I make sure I get one with a "good" BIOS?

[bigbrovar (bigbrovar)]

Seriously guys the more "me too comments "you add to this post, the more
distracted the developers are, please this bug is being worked on both
upstream and downstream. Coming here to express you pain would not help
solve this any further just add more distraction and frustrate the people
working on this. If you have any new information that might help in fixing
this by all means add. If all you have is make sentimental statements please
use the forums. If this bug affects you. Indicate using the appropriate
tool.
On Jul 18, 2011 6:06 PM, "Derek P" <email address hidden> wrote:
> Novice user adding input here. If this cannot be fixed so that I can
> have similar battery life as when running Windows I will have to stop
> using Ubuntu. I prefer this OS over Windows 7 but I can't work with
> barely 2 hours of battery life. This is with -
> GRUB_CMDLINE_LINUX_DEFAULT="quiet splash pcie_aspm=force"
>
> I am crossing my fingers they figure something out. Thanks for
> listening.
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/760131
>
> Title:
> Power consumption raised significantly in natty
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/ubuntu-release-notes/+bug/760131/+subscriptions

[Richard Kleeman (kleeman)]

Well actually according to Jonathan Corbet editor of Linux Weekly News (and a kernel expert) the problem has been all along that this has not been properly brought to the attention of the kernel developers which is where the solution surely lies.

This could be achieved by a kernel literate Ubuntu user or better still a Ubuntu kernel developer lodging a detailed bug report on the main Kernel bugzilla.

I would have done this but do not feel literate enough (and do not have the time) to give the relevant info needed by Linus and team.

If this has been done already please disregard but point me to where exactly.

[Nicolas Michel (nicolas-michel)]

The problem is that we don't know if someone is working on it since both
articles I read on that topic (phoronix, lwn) says that nobody seems to
take care of it.

On 07/19/2011 08:44 PM, bigbrovar wrote:
> Seriously guys the more "me too comments "you add to this post, the more
> distracted the developers are, please this bug is being worked on both
> upstream and downstream. Coming here to express you pain would not help
> solve this any further just add more distraction and frustrate the people
> working on this. If you have any new information that might help in fixing
> this by all means add. If all you have is make sentimental statements please
> use the forums. If this bug affects you. Indicate using the appropriate
> tool.
> On Jul 18, 2011 6:06 PM, "Derek P"<email address hidden> wrote:
>> Novice user adding input here. If this cannot be fixed so that I can
>> have similar battery life as when running Windows I will have to stop
>> using Ubuntu. I prefer this OS over Windows 7 but I can't work with
>> barely 2 hours of battery life. This is with -
>> GRUB_CMDLINE_LINUX_DEFAULT="quiet splash pcie_aspm=force"
>>
>> I am crossing my fingers they figure something out. Thanks for
>> listening.
>>
>> --
>> You received this bug notification because you are subscribed to the bug
>> report.
>> https://bugs.launchpad.net/bugs/760131
>>
>> Title:
>> Power consumption raised significantly in natty
>>
>> To manage notifications about this bug go to:
>> https://bugs.launchpad.net/ubuntu-release-notes/+bug/760131/+subscriptions
>

[Tamran (tamran-lengyel)]

I read all of the posts here and a few others elsewhere. I tried a few things:

1) I tried the 'pcie_aspm=force' trick. It honestly did nothing for me. I worked with it for a couple of days. I see some people are saying that it helped, so who knows.

2) I installed the kernel from this link: http://kernel.ubuntu.com/~kernel-ppa/mainline/v2.6.34.7-maverick/
I downloaded the following files (for a 64bit system) into a directory:

linux-headers-2.6.34-02063407-generic_2.6.34-02063407.201009140905_amd64.deb
linux-headers-2.6.34-02063407_2.6.34-02063407.201009140905_all.deb
linux-image-2.6.34-02063407-generic_2.6.34-02063407.201009140905_amd64.deb

Then I went to a console and went into that directory and typed:

> sudo dpkg -i *.deb

When I rebooted, I chose the 2.3.34 kernel. I've been running with this configuration for over a day now and I played angry birds for 2 hours straight without an overheat/shutdown (it took me 5-10 min for a shutdown before reverting the kernel).

I know this is a little bit on the "not supported" side, but this is the only thing that worked for me. I chose the 2.6.34 kernel based on what was said here and in some forums posts.

Best of luck,

Tamran

[lokster (lokiisyourmaster)]

@bigbrovar "Seriously guys the more "me too comments "you add to this post, the more
distracted the developers are"
---
Are you kidding? I don't see them working on this, so how distracted they can get?!
I don't know if they are working on fixing this or not, but my impression is that they don't really care.
Please, prove me wrong.
And yes - I will say "me too". The 'pcie_aspm=force' helped on my netbook - it's cooler now, and consumes approx. 1W less.

[Colin Ian King (colin-king)]

I suggest testing out the latest natty -proposed kernel, it contains commit 9ee653dce0efc6bad29f0d68b4ac74dbed093131

commit 9ee653dce0efc6bad29f0d68b4ac74dbed093131
Author: Tero Kristo <email address hidden>
Date: Thu Feb 24 17:19:23 2011 +0200

    cpuidle: menu: fixed wrapping timers at 4.294 seconds

    BugLink: http://bugs.launchpad.net/bugs/774947

    Cpuidle menu governor is using u32 as a temporary datatype for storing
    nanosecond values which wrap around at 4.294 seconds. This causes errors
    in predicted sleep times resulting in higher than should be C state
    selection and increased power consumption. This also breaks cpuidle
    state residency statistics.

    cc: <email address hidden> # .32.x through .39.x
    Signed-off-by: Tero Kristo <email address hidden>
    Signed-off-by: Len Brown <email address hidden>
    (cherry picked from commit 7467571f4480b273007517b26297c07154c73924)
    Acked-by: Stefan Bader <email address hidden>
    Signed-off-by: Tim Gardner <email address hidden>

This should fix a bug with incorrect C state selection which lead to increased power consumption.

[alizard (alizard)]

if you page to the end of the posts on the bug you reference, you will find:
---------- quote
On 07/20/2011 03:24 AM,
Colin King wrote:Chris Van Hoof (vanhoof) wrote on 2011-07-13:
#34 Marking as fix released as 2.6.38-10 was released today

Changed in linux (Ubuntu Natty):
status: Fix Committed → Fix Released
---------end quote -------------------------------
Another post in that discussion says the fix was already pushed out to
Oneric.

alizard@terrarium:~$ uname -a
Linux terrarium 2.6.38-10-generic-pae #46-Ubuntu SMP Tue Jun 28 16:54:49
UTC 2011 i686 athlon i386 GNU/Linux
alizard@terrarium:~$

My desktop runs Natty, got the regular update upgrading the kernel to
38-10 and appears to be running several degrees cooler. Did not observe
this in detail, as I was far more concerned about my netbook which is
running 5 C hotter than usual.

My netbook will be upgraded ASAP, the rest of you might want to try
upgrading to a current kernel version and see if you still have the problem.

YMMV and good luck.
A.Lizard

> I suggest testing out the latest natty -proposed kernel, it contains
> commit 9ee653dce0efc6bad29f0d68b4ac74dbed093131
>
> commit 9ee653dce0efc6bad29f0d68b4ac74dbed093131
> Author: Tero Kristo <email address hidden>
> Date: Thu Feb 24 17:19:23 2011 +0200
>
> cpuidle: menu: fixed wrapping timers at 4.294 seconds
>
> BugLink: http://bugs.launchpad.net/bugs/774947
>
> Cpuidle menu governor is using u32 as a temporary datatype for storing
> nanosecond values which wrap around at 4.294 seconds. This causes errors
> in predicted sleep times resulting in higher than should be C state
> selection and increased power consumption. This also breaks cpuidle
> state residency statistics.
>
> cc: <email address hidden> # .32.x through .39.x
> Signed-off-by: Tero Kristo <email address hidden>
> Signed-off-by: Len Brown <email address hidden>
> (cherry picked from commit 7467571f4480b273007517b26297c07154c73924)
> Acked-by: Stefan Bader <email address hidden>
> Signed-off-by: Tim Gardner <email address hidden>
>
> This should fix a bug with incorrect C state selection which lead to
> increased power consumption.
>
>

--
member The Internet Society (ISOC), The HTML Writers Guild.
"All that is, is information"
Personal Website http://www.ecis.com/~alizard
business Website http://www.reptilelabs.com
backup address (if ALL else fails) <email address hidden>
PGP 8.0 key available by request or keyserver. Download PGP from:
http://www.pgpi.org for e-mail privacy.
Disaster prep info: http://www.ecis.com/~alizard/y2k.html
***Looking for INTELLIGENT new technology public policy alternatives?***
http://www.ecis.com/~alizard/technology.html

[Erwin Junge (erwin-junge)]

I did some power tests on my Samsung N220 Plus (atom n450 based netbook) using phoronix test suite battery power usage test.

The details are published here: http://openbenchmarking.org/result/1107249-GR-1107245GR43

In summary, the results are this (average power use in mW):
10.10 7660.37
11.04 unity pcie_aspm=force 8263.59
11.04 classic (no effects) pcie_aspm=force 8156.98
11.04 unity 8795.90
11.04 classic (no effects) 8194.49

Especially notice the following points:
- Difference between unity and classic (no effects) when not using pcie_aspm=force --> ~600 mW (~7%)
- Difference between unity and classic (no effects) when using pcie_aspm=force --> ~100 mW (~1%)
- Lack of difference between pcie_aspm=force and regular when using classic (no effects) --> ~40 mW (~0.5%)
- Difference between 10.10 and 11.04 classic (no effects) with or without pcie_aspm=force --> ~500 mW (~7%)

Notes: 10.10 was using no effects, desktop edition ubuntu (so no compiz)

Conclusions:
- pcie_aspm=force makes a difference when using unity
- pcie_aspm=force makes no difference when not using desktop effects
- when not using desktop effects, 11.04 still takes 500 mW more than 10.10 (independent of pcie_aspm=force)

So pcie_aspm=force does not fix the entire problem and (at least in my tests) only makes a difference when unity/compiz is used. This might be related to gpu usage for desktop rendering, but I don't know enough about the differences to really comment on this. If anyone knows more tests I can do to further narrow down what might be the cause, please let me know.

Greetings,

Erwin Junge

[Erwin Junge (erwin-junge)]

I upgraded to 2.6.38-11 and ran the battery-power-usage test from phoronix test suite.

The results have been added to: http://openbenchmarking.org/result/1107243-GR-1107245GR40

The summary is as follows (avg power usage in mW):

unity aspm 8582.55
classic aspm 8402.90
unity 8184.63
classic 8150.33

classic is again the (no effects) variety.

Oddly, the latest kernel update has made pcie_aspm=force increase power usage vs no pcie_aspm=force. I'm completely stumped. Also, the difference between unity or classic has vanished when not using pcie_aspm=force. The difference of ~500 mW wrt 10.10 is still there though.

I'm getting the idea that pcie_aspm is a lot less related to the current power issues than we previously thought. Any ideas on extra tests to either confirm or deny this are very welcome and will be executed swiftly. I have two of these machines (identical), so can potentially perform different tests in parallel.

Greetings,

Erwin Junge

[Peter Kornhuld (webmaaschter)]

Hi,

On ubuntu 11.04, Acer eeePC, I included this into /etc/rc.local to reduce power consumption by ~30%.

----------
for F in /sys/bus/usb/devices/*/power/control; do
 echo "auto" > $F
done
----------

Greetings
Peter

[Matthew Hessel (matt-hessel)]

@Erwin - not sure if you have this same hardware as mine, but I did find what makes the difference for my laptop.

Natty running 2.6.35 kernel, can idle at 18.5 watts - runs ok, but I find that power management screws up my primary drive when it goes to sleep, will not wake up the drive at all, system just locks hard - on Seagate Hybrid drive. On Western Digital drive - no issues with wake up. realtek based intel-hda shows ASPM unknown and disabled

sudo lspci -vvv shows that most things run with ASPM enabled, iwlagn, usb ports, half of the PCIE ports, and the video (Nvidia-running binary blob)

Natty running default kernel 2.6.38 - Even with pcie_aspm=force, iwlagn and the pcie root it sits on has ASPM disabled for some reason. intel-hda still not supporting ASPM - Idle with powertop shows my system it at 21.5 watts.

Natty running current upstream oneiric builds, 3.0.0.8. ASPM is the same, the kernel tweaks found after 2.6.38 with the idle changes helps some, best idle level is about 20 watts

Natty running upstream vanilla from git.kernel.org, ASPM works for the intel-hda card. iwlagn still is not enabling ASPM. Idle is at 19.8 watts

found good info at

http://linuxwireless.org/en/users/Documentation/ASPM

using that page, I looked at the output from LSPCI for the wireless card and the PCIe root for it, found that the bits were not set for aspm (hex= 0x40) I unloaded the iwlagn driver and changed to 0x43 for the root and the iwlagn stuff, then modprobe iwlagn again..

ASPM then shows it is working, - and confirm it works with the 2.6.38 kernel all the way up to the upstream sources - 3.0.0.8 and the vanilla source.

with the oneiric kernel, I seem to get the best results, 17.8 for idle wattage.. better than the maverick sources.

the pcie_aspm=force is only a workaround needed if your bios isn't set up right. if you run 'dmesg | grep OSC' I see

[ 2.061987] pci0000:00: Requesting ACPI _OSC control (0x1d)
[ 2.061992] pci0000:00: ACPI _OSC request failed (AE_NOT_FOUND), returned control mask: 0x1d
[ 2.061995] ACPI _OSC control for PCIe not granted, disabling ASPM

on the laptop, where it helps, but on another desktop --

[ 0.716560] pci0000:00: Requesting ACPI _OSC control (0x1d)
[ 0.717137] pci0000:00: ACPI _OSC control (0x15) granted

where the pcie_aspm=force does nothing at all..

[Erwin Junge (erwin-junge)]

@Matthew

What I understood from your post is that there was a bug in your wireless driver and you manually set the aspm bit which resulted in a significant power reduction?

Sounds very dangerous to me (at least the stuff I read in that wiki sounds way too prone to user error), but thanks for posting the info. Maybe this could be used by someone more skilled in this than me as the start of a workaround?

[Matthew Hessel (matt-hessel)]

Yes, you got the right idea.

I don't know that it is really dangerous - worst case would be your
driver wouldn't load or lock up, a reboot would fix it.

My scenario works for the iwlagn driver, I don't know if you have the
same hardware.

It does make a 2 - 3 watt difference in power usage for me.

I don't know why this changed as it worked automatically in the 2.6.35
tree, but not now. I have a script manually perform the setpci mods
before the kernel loads the iwlagn driver which takes care of it for
now.

On a different note, there have been a few posts for eepc's in the
thread, the natty kernel didn't have the WMI modules enabled for
them. Maverick did, there was a commit in the oneiric tree that put
them back in - that should correct that issue.

If you run one of those, try the current kernel, I think it is 3.0.0-8
right now.

On 8/11/11, Erwin Junge <email address hidden> wrote:
> @Matthew
>
> What I understood from your post is that there was a bug in your
> wireless driver and you manually set the aspm bit which resulted in a
> significant power reduction?
>
> Sounds very dangerous to me (at least the stuff I read in that wiki
> sounds way too prone to user error), but thanks for posting the info.
> Maybe this could be used by someone more skilled in this than me as the
> start of a workaround?
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/760131
>
> Title:
> Power consumption raised significantly in natty
>
> Status in Release Notes for Ubuntu:
> Fix Released
> Status in “linux” package in Ubuntu:
> Triaged
> Status in “linux” source package in Natty:
> Incomplete
> Status in “linux” source package in Oneiric:
> Triaged
>
> Bug description:
> This is a regression moving from maverick->natty. With the same sort
> of light interactive usage, Battery life has gone from ~4 hours to
> 2:45 or so.
>
> Running powertop shows the system in P-state "Turbo mode" 10-15% of
> the time. Typically this was almost zero in maverick, 98+% of the
> time maverick was in the lowest state.
>
> wakeups reported in the 5-600/s range where previously (from memory)
> it was 200 ish. i915 driver is high on the list. This is
> irrespective of whether running unity or classic desktop.
>
> ProblemType: Bug
> DistroRelease: Ubuntu 11.04
> Package: linux-image-2.6.38-8-generic 2.6.38-8.42
> Regression: Yes
> Reproducible: Yes
> ProcVersionSignature: Ubuntu 2.6.38-8.42-generic 2.6.38.2
> Uname: Linux 2.6.38-8-generic x86_64
> AlsaVersion: Advanced Linux Sound Architecture Driver Version 1.0.23.
> Architecture: amd64
> ArecordDevices:
>  **** List of CAPTURE Hardware Devices ****
>  card 0: Intel [HDA Intel], device 0: CONEXANT Analog [CONEXANT Analog]
>    Subdevices: 1/1
>    Subdevice #0: subdevice #0
> AudioDevicesInUse:
>  USER PID ACCESS COMMAND
>  /dev/snd/controlC0: james 2068 F.... pulseaudio
> CRDA: Error: [Errno 2] No such file or directory
> Card0.Amixer.info:
>  Card hw:0 'Intel'/'HDA Intel at 0xf2620000 irq 47'
>    Mixer name : 'Conexant CX20561 (Hermosa)'
>    Components : 'HDA:14f1...

[Tamran (tamran-lengyel)]

@Matthew: Are you by any chance using the following intel wireless card?

Intel Corporation PRO/Wireless 5100 AGN [Shiloh]

One thing I have not seen mention until now, but have noticed is that my laptop is REALLY hot near my wireless card. Even with the 2.6.34 kernel (which I've been using for a few weeks).

Tamran

[Matthew Hessel (matt-hessel)]

Yes, I have the 5100

On 8/12/11, Tamran <email address hidden> wrote:
> @Matthew: Are you by any chance using the following intel wireless card?
>
> Intel Corporation PRO/Wireless 5100 AGN [Shiloh]
>
> One thing I have not seen mention until now, but have noticed is that my
> laptop is REALLY hot near my wireless card. Even with the 2.6.34 kernel
> (which I've been using for a few weeks).
>
> Tamran
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/760131
>
> Title:
> Power consumption raised significantly in natty
>
> Status in Release Notes for Ubuntu:
> Fix Released
> Status in “linux” package in Ubuntu:
> Triaged
> Status in “linux” source package in Natty:
> Incomplete
> Status in “linux” source package in Oneiric:
> Triaged
>
> Bug description:
> This is a regression moving from maverick->natty. With the same sort
> of light interactive usage, Battery life has gone from ~4 hours to
> 2:45 or so.
>
> Running powertop shows the system in P-state "Turbo mode" 10-15% of
> the time. Typically this was almost zero in maverick, 98+% of the
> time maverick was in the lowest state.
>
> wakeups reported in the 5-600/s range where previously (from memory)
> it was 200 ish. i915 driver is high on the list. This is
> irrespective of whether running unity or classic desktop.
>
> ProblemType: Bug
> DistroRelease: Ubuntu 11.04
> Package: linux-image-2.6.38-8-generic 2.6.38-8.42
> Regression: Yes
> Reproducible: Yes
> ProcVersionSignature: Ubuntu 2.6.38-8.42-generic 2.6.38.2
> Uname: Linux 2.6.38-8-generic x86_64
> AlsaVersion: Advanced Linux Sound Architecture Driver Version 1.0.23.
> Architecture: amd64
> ArecordDevices:
>  **** List of CAPTURE Hardware Devices ****
>  card 0: Intel [HDA Intel], device 0: CONEXANT Analog [CONEXANT Analog]
>    Subdevices: 1/1
>    Subdevice #0: subdevice #0
> AudioDevicesInUse:
>  USER PID ACCESS COMMAND
>  /dev/snd/controlC0: james 2068 F.... pulseaudio
> CRDA: Error: [Errno 2] No such file or directory
> Card0.Amixer.info:
>  Card hw:0 'Intel'/'HDA Intel at 0xf2620000 irq 47'
>    Mixer name : 'Conexant CX20561 (Hermosa)'
>    Components : 'HDA:14f15051,17aa20ff,00100000'
>    Controls : 16
>    Simple ctrls : 8
> Card29.Amixer.info:
>  Card hw:29 'ThinkPadEC'/'ThinkPad Console Audio Control at EC reg 0x30,
> fw 7XHT24WW-1.06'
>    Mixer name : 'ThinkPad EC 7XHT24WW-1.06'
>    Components : ''
>    Controls : 1
>    Simple ctrls : 1
> Card29.Amixer.values:
>  Simple mixer control 'Console',0
>    Capabilities: pswitch pswitch-joined penum
>    Playback channels: Mono
>    Mono: Playback [on]
> Date: Wed Apr 13 15:03:35 2011
> EcryptfsInUse: Yes
> HibernationDevice: RESUME=UUID=bb479652-f524-4abe-b1b0-27646d6deebc
> InstallationMedia: Ubuntu 10.10 "Maverick Meerkat" - Release Candidate
> amd64 (20100928)
> MachineType: LENOVO 7465CTO
> ProcEnviron:
>  LANGUAGE=en_US:en_GB:en
>  PATH=(custom, user)
>  LANG=en_US.UTF-8
>  SHELL=/bin/bash
> ProcKernelCmdLine: BOOT_IMAGE=/boot/vmlinuz-2.6.38-8-...

[rockachu2 (rockachu2)]

I seem to remember a kernel patch that increased speed and responsiveness being added recently. Would this be the cause? cause then it isn't a regression, just a utilization of more power to provide more responsiveness.

[fejes (anthony-fejes)]

Even if that were so, when my laptop goes into battery mode, with all of the aggressive power saving modes throttling the CPU, etc, it's unacceptable for it to be running so hot. I wouldn't normally care if it's plugged in and sucking up power, but when the power is supposed to be scaled back, it shouldn't be taking up more power ( even at the cost of decreased responsiveness.)

In any case, as of yesterday, after a kernel update, my laptop is now consuming 14-17W of power when on battery, as opposed to the 19-25W it was taking last week. While not as good as the 11-14W range I had going with Maverick, it's an improvement. I don't recall seeing anything in the kernel changelog that would indicate a change to the power consumption, but I'm not a kernel expert. In any case, I have not altered any of my power saving optimizations between the two tests, and I had used the force acpi trick much earlier on.

If it stays this way for now, that's not a bad sign.

[Ganesh S Hegde (ganumama)]

i am running 3.0.3 compiled from source in natty , and not
experiencing any battery or heat issues... i get 11 -14 w max
consumption in ... and upto 19-20 W on AC,i have kept cpu freq to
'ondemand' and also i have turned on "pcie.aspm=force", I guess the
issue is fixed in 3.0 kernel series.... i dint apply any patches, or
tweaks, just did a simple menuconfig.
the most important thing is laptop doesnt heat now (60 degrees in
2.6.37 v/s 55 in 3.0.3)
mine is a HP g42, Core i3-370M, 3GB DDR3,
<hopefully i dint troll>

On 9/2/11, fejes <email address hidden> wrote:
> Even if that were so, when my laptop goes into battery mode, with all of
> the aggressive power saving modes throttling the CPU, etc, it's
> unacceptable for it to be running so hot. I wouldn't normally care if
> it's plugged in and sucking up power, but when the power is supposed to
> be scaled back, it shouldn't be taking up more power ( even at the cost
> of decreased responsiveness.)
>
> In any case, as of yesterday, after a kernel update, my laptop is now
> consuming 14-17W of power when on battery, as opposed to the 19-25W it
> was taking last week. While not as good as the 11-14W range I had going
> with Maverick, it's an improvement. I don't recall seeing anything in
> the kernel changelog that would indicate a change to the power
> consumption, but I'm not a kernel expert. In any case, I have not
> altered any of my power saving optimizations between the two tests, and
> I had used the force acpi trick much earlier on.
>
> If it stays this way for now, that's not a bad sign.
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/760131
>
> Title:
> Power consumption raised significantly in natty
>
> Status in Release Notes for Ubuntu:
> Fix Released
> Status in “linux” package in Ubuntu:
> Triaged
> Status in “linux” source package in Natty:
> Incomplete
> Status in “linux” source package in Oneiric:
> Triaged
>
> Bug description:
> This is a regression moving from maverick->natty. With the same sort
> of light interactive usage, Battery life has gone from ~4 hours to
> 2:45 or so.
>
> Running powertop shows the system in P-state "Turbo mode" 10-15% of
> the time. Typically this was almost zero in maverick, 98+% of the
> time maverick was in the lowest state.
>
> wakeups reported in the 5-600/s range where previously (from memory)
> it was 200 ish. i915 driver is high on the list. This is
> irrespective of whether running unity or classic desktop.
>
> ProblemType: Bug
> DistroRelease: Ubuntu 11.04
> Package: linux-image-2.6.38-8-generic 2.6.38-8.42
> Regression: Yes
> Reproducible: Yes
> ProcVersionSignature: Ubuntu 2.6.38-8.42-generic 2.6.38.2
> Uname: Linux 2.6.38-8-generic x86_64
> AlsaVersion: Advanced Linux Sound Architecture Driver Version 1.0.23.
> Architecture: amd64
> ArecordDevices:
>  **** List of CAPTURE Hardware Devices ****
>  card 0: Intel [HDA Intel], device 0: CONEXANT Analog [CONEXANT Analog]
>    Subdevices: 1/1
>    Subdevice #0: subdevice #0
> AudioDevicesInUse:
>  USER PID ACCESS COMMAND
>  /dev/snd/cont...

[fejes (anthony-fejes)]

I should probably have mentioned that I'm running Oneric, not Natty, although I'm not certain it's fixed - I just wanted to note that the 3.0.0.9 kernel seems to have made an improvement.

[Tim Gardner (timg-tpi)]

SRU Justification:

Impact: Since 2.6.36 (23016bf0d25), Linux prints the existence of "epb" in /proc/cpuinfo,
    Since 2.6.38 (d5532ee7b40), the x86_energy_perf_policy(8) utility has
    been available in-tree to update MSR_IA32_ENERGY_PERF_BIAS.

    However, the typical BIOS fails to initialize the MSR, presumably
    because this is handled by high-volume shrink-wrap operating systems...

    Linux distros, on the other hand, do not yet invoke x86_energy_perf_policy(8).
    As a result, WSM-EP, SNB, and later hardware from Intel will run in its
    default hardware power-on state (performance), which assumes that users
    care for performance at all costs and not for energy efficiency.
    While that is fine for performance benchmarks, the hardware's intended default
    operating point is "normal" mode...

    Initialize the MSR to the "normal" by default during kernel boot.

    x86_energy_perf_policy(8) is available to change the default after boot,
    should the user have a different preference.

Patch: commit abe48b108247e9b90b4c6739662a2e5c765ed114 upstream

[bas india (baskarans)]

Hi Anyone tested this fix? Awaiting for result. Anyway thanks for Cannonical kernel team!!! - Bas

[florin (florin-arjocu)]

Was it pushed to ppa? I just checked and I don't have any kernel updates today or yesterday.

[Tim Gardner (timg-tpi)]

florin - tip of the tree builds are performed daily and can be found here: https://launchpad.net/~kernel-ppa/+archive/pre-proposed

[ronexus (ronexus)]

Whooo!
With the latest package-upgrade (i.e. linux-image 2.6.38-11) for Natty, the
Idle Power consumption reported by powertop on my TP T410 went down *more
than 50%*: From around 25W to below 12W.
It is also much cool and quiet again.

Just awesome, thank you all so much for the great work on this issue :)

[Bartosz Zasieczny (siekacz)]

When this patch will be applicable in oneiric? My Vostro 3450 eats whole battery in 1.5h, and ASPM workaround doesn't work.

[Carlos Moffat (carlos-eldiabloenlosdetalles)]

I have been using the latest stable kernels from "mainline" on my X220, but I was curious to see if the power consumption was better with the 'pre-proposed' kernel mentioned in #167. For some reason, however, the computer would start but it would hang just after login in gnome (the panel wouldn't show up), so I had to go back to 3.0.4.

Is this patch going to make it to those mainline kernels?

Cheers

[George (geoaraujo)]

I've installed it with ppa:kernel-ppa/pre-proposed
But my laptop now seems to run hotter than before... :(

[Kevin (kevinshlee)]

I've upgraded to 2.6.38-11.51~pre201109220902 in the kernel-ppa/pre-proposed. Unfortunately, I don't think it works on my Lenovo E520. The battery life shown is about 2 hours then it drains so fast while I can have about 5 hours on Windows 7.

[Kevin (kevinshlee)]

By the way, I'm using Ubuntu 11.04 desktop 64bit.

[Kevin (kevinshlee)]

In my case, pcie_aspm=force doesn't seem to work either. I tried it after having kernel upgrade with the latest one from kernel-ppa/pre-proposed and only through command-line modification when booting. I can see "PCIe ASPM is forcedly enabled" in /var/log/dmesg but don't see any difference.

[George (geoaraujo)]

By the way, how do I uninstall the kernel ppa:kernel-ppa/pre-proposed installed and get the old one back?

[Philip Khor (philip10khor)]

I really hope a fix will be committed as soon as possible in Oneiric. My laptop's battery life is terrible and I get a prompt that I have approximately 20 minutes left at 40%. The computer hibernates about 5 minutes later, even while using pcie_aspm=force. I'm using kernel 3.0.0-11-generic.

[Philip Muškovac (yofel)]

Please don't change the Bug Status randomly without commenting why. Also don't touch a kernel bug's status unless you know what the states mean.

[Leann Ogasawara (leannogasawara)]

The patch noted in comment #164 was applied to Oneiric when the Oneiric kernel was rebased to upstream stable v3.0.1. It was subsequently uploaded and available in Oneiric as of linux-3.0.0-8.10. I'm therefore marking the Oneiric task Fix Released. If you are still experiencing issues with regards to power consumption even with this patch applied, I strongly urge you to open a new bug report as issues such as this are typically hardware specific and will require different fixes. Thanks.

[Kevin (kevinshlee)]

I do still have this issue with the kernel 2.6.38-11.51~pre201109230902 from https://launchpad.net/~kernel-ppa/+archive/pre-proposed.
Please see
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/760131/comments/172
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/760131/comments/173
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/760131/comments/174

Should I create a new ticket for this although I'm not using Oneiric but Natty?
If so, could you please let me know what kind of information I should provide to help the kernel team solve this issue? I'm so eager to help as I really don't want to use Windows 7. It's so inconvenient to use. Please help us use Ubuntu again.

[Deltik (deltik)]

The fix hasn't been released for Natty. While you're using Natty, you'll have to wait for a release.

[Kevin (kevinshlee)]

Thanks Deltik, but as I said, I was using 2.6.38-11.51~pre201109230902 from https://launchpad.net/~kernel-ppa/+archive/pre-proposed when I put the comment.
Wasn't it the version to which the patch was applied? Please correct me if I'm wrong.
Anyway, I now have the 2.6.38-12 one from https://launchpad.net/~kernel-ppa/+archive/pre-proposed and still have the same problem.

If I'm wrong, could anyone please tell me where I can get the one with the patch applied? I mean for natty.

[madbiologist (me-again)]

Kevin - a check of the publishing details shows that the patch noted in comment #164 is included in both the kernels you mentioned in comment #181.

What CPU is in your Lenovo E520?

[bas india (baskarans)]

I too having same issue even after upgraded the kernel version from 2.6.38-11.48 to 2.6.38-11.50.
My system is Dell Inspiron1564 | Intel i3 Processor | 2GB RAM | Ubuntu Natty

Is the patch is system specific?

Awaiting for the patch since its affects my laptop battery life significantly.

Thanks in advance.

[Kevin (kevinshlee)]

Thanks madbiologist for your answer.

It's Intel i7 2620M. It has an AMD Radeon HD 6630M switchable graphics chip but I turned it off by following http://ubuntuforums.org/showthread.php?t=1752202
More specifically http://ubuntuforums.org/showpost.php?s=cf99e4261d426e3add7cb3413d0fd8f1&p=10832728&postcount=7 and http://ubuntuforums.org/showpost.php?p=10803370&postcount=27

I tested with Oneiric and its latest kernel but it didn't work. With Oneiric, it seems to drain the battery even faster.
I only tested it with Live USB but updated it to get the latest kernel and other software packages.

For now, I've completely got rid of Ubuntu and am using Windows 7 which I really didn't want. On Windows 7, I have about 5+ hour battery life whereas with Natty, it's about 2 hours (with Oneiric, it seems shorter).

[Ganesh S Hegde (ganumama)]

i have not tested the patched kernel in natty but as far as oneiric goes the power regression is significantly higher than it was ever in Natty... i see 22 W idle and upto 34 W usage on normal browsing in oneiric 3.0.0.11 which is the Oneiric Beta 2 kernel... it was mentioned in #179 that the fix has been released, but despite having a first generation core i3-370M processor. I cant report a duplicate bug about the same issue . Please fix it in Oneiric or atleast keep a seperate older kernel based on 2.6.37.6 or earlier seperately for laptop users.....

[Bartosz Zasieczny (siekacz)]

On my Dell Vostro 3450 with i5 2410M Ubuntu can work over 4 hours without any power supply. I just had to force ASPM to work and put "echo powersave > /sys/module/pcie_aspm/parameters/policy" in /etc/rc.local. I know it's an awful hack, but it works. Something should be done about it. Battery lifetime should be maximized out of the box.

[Herton R. Krzesinski (herton)]

This bug is awaiting verification that the Natty kernel in -proposed solves the problem.

Please test the following way:
- Test the current 2.6.38-11.50 kernel in -updates. The problem should be present.
- Then update to the 2.6.38-12.51 kernel currently in -proposed. If the problem is solved, change the tag 'verification-needed-natty' in this bug to 'verification-done-natty'. If this verification is not done by one week from today, the applied fix from comment #164 will be dropped from the source code, and this bug will be closed. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you!

If you have a power consumption which is good/expected with 2.6.38-11.50 kernel, and it's bad with 2.6.38-12.51 kernel, likely there is a regression, in this case please report here.

Regressions from another natty versions should be in a separate bug report.

In case none of the kernels since natty release had good/expected power consumption on your machine, likely you have a different bug from this one (typically hardware specific requiring different fixes). It's hard to isolate problems with all mixed reports here. In this case please open a new bug as already requested.

[Dragoneyes (stiu)]

The kernel 2.6.38-12.51 don't solves the problem for me. I see a little improvement in battery life (from 4h to 4h30'), but powertop doesn't notice any reduction in power usage in idle. Moreover, the optimal battery life of my netbook should be about 6h. ( aspm parameter at boot does nothing).
I'm with a Zareason Terra HD (Intel Atom N450 and Wifi Link 5100).
I'm going to open a new bug a suggested, but with the oncoming release of Oneiric in a few days.

[mejo (jonas-freesources)]

Unfortunately kernel 2.6.38-12.51 doesn't improve the situation that much for me either. Maybe a bit, but I'm not sure about that. I was surprised to find out that even pcie_aspm=force and i915_enable_rc6=1 don't improve power consumption anymore. Some months back pcie_aspm=force still improved the situation for me.

My system is a ThinkPad T420 with Sandy Bridge i5, integrated i915 video controller and running up-to-date Ubuntu Natty.

I compared the output of powertop with kernel 2.6.38-11.50 and with kernel 2.6.38-12.51, both with and without the boot options, and with and without firefox running. See my exact results attached.

[Michael Larabel (michael-phoronix)]

Here's my findings on the latest power developments, for what it's worth: http://www.phoronix.com/vr.php?view=16496

[Suhel (yours-rey)]

I've recently installed 11.10 beta 2 on my laptop {Dell Xps 15, i5, 6gbram, 1gb video card, 500gb hd) and its sucking my battery like a sponge would suck water I hardly get an hour of battery backup. whereas when I use Windows7 my battery lasts more than 2:30hours.
I've the latest kernels upgraded and its not fixed. (who marked this bug fixed?????)

[Pako (elektrobank01)]

There are many reports from users about high xorg CPU load when using themes like: Ambiance, Radiance, Dust, Human and Impression. Clearlooks is the only theme which consumes lowest power.

One of the reported bug: Bug #595845

[vervelover (alessiopangos)]

Apparently, there is no fix for this bug yet, proof is here:

http://www.phoronix.com/scan.php?page=article&item=ubuntu_linux_epb&num=1

Quoting the article:

"What Canonical's kernel team says is the "fix" to LP #760131 is an unrelated power fix done upstream to ensure that the MSR_IA32_ENERGY_PERF_BIAS register is set to "normal" rather than "performance" mode. The hardware in the original posting to the aforementioned bug report does not even support EPB and is about ASPM and other power changes causing the significantly higher power usage. As these latest tests from the Sandy Bridge notebook show, the 3.0.1+ kernels can improve the power efficiency without degrading the system performance, but will only help you if running the latest Intel hardware that supports the energy performance bias feature."

I would suggest not to release Oneiric at all, you can't be taken seriously with an OS eating laptop batteries twice as fast as it should, no other non-linux OS was ever released with a bug like this.

But please, at least do not mark this as "fix released".

[Pako (elektrobank01)]

Can someone test the same kernel on Xubuntu or Lubuntu, because It's а little bit suspicious on how Ubuntu draw and render the GUI, just to be sure that everything is ok there, plus, there are almost 20 power hungry non-kernel services running in background. Test!

[fejes (anthony-fejes)]

Just to add a counterpoint, I've found that things are more or less working well for me with the new kernel - it's definitely better then the earlier versions. However, my biggest problem is the brcmsmac driver being pushed out without power management, which appears to be taking up to 5W (5W!) of power, and never less than (3.5W). I realize it's an unrelated bug, (I've filed it here: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/867869), but the kernel is averaging 9-11W for the most part - which isn't too bad.

I just suspect power management has not been a priority for Oneiric.

[Erwin Junge (erwin-junge)]

I am now running Lubuntu Beta 2 with kernel 3.0.0-12-generic and at least for me the issue seems fixed. I didn't do any actual wattage measurements, but my time on battery went from ~6 hours to ~10 hours (where it was before this whole power problem started).

Since this problem seems to be very hardware specific, I am running a netbook with and Intel Atom N450 (CPU + IGP). No seperate videocard.

This bug probably should be split up into multiple other bugs for the people still experiencing problems, with the split being based on hardware characteristics. Reading through all of the comments suggests that this has become a melting pot of multiple different issues, which is counterproductive to the actual finding and fixing of said bugs.

[Pako (elektrobank01)]

@ Erwin Junge

How you get your result? ~6 hours in Ubuntu, ~10 in Lubuntu? or ~6 to ~10 in Lubuntu with a different kernel?
However, Erwins test shows that there are huge difference in battery time from different desktop environments, thus, you can't get results based on kernel test only, because it's just a fraction from hundrets of other consumers, including: cairo, compiz, metacity, pixbuf, xorg, mesa, wifi, open-closed video drivers, Mutter, Unity, indicator applets... etc.
Can someone test the latest kernel on Ubuntu 8.10?

[Erwin Junge (erwin-junge)]

On Mon, 10 Oct 2011 10:29:21 -0000, Pako <email address hidden>
wrote:
> @ Erwin Junge
>
> How you get your result? ~6 hours in Ubuntu, ~10 in Lubuntu? or ~6 to
~10
> in Lubuntu with a different kernel?
> However, Erwins test shows that there are huge difference in battery
time
> from different desktop environments, thus, you can't get results based
on
> kernel test only, because it's just a fraction from hundrets of other
> consumers, including: cairo, compiz, metacity, pixbuf, xorg, mesa, wifi,
> open-closed video drivers, Mutter, Unity, indicator applets... etc.
> Can someone test the latest kernel on Ubuntu 8.10?

~6 hours in Ubuntu 11.04, ~10 hours in Lubuntu 11.10. I agree that the
cause can't be known from this, since too many things changed all at once
going from Ubuntu 11.04 to Lubuntu 11.10. I don't have time to do more
tests right now, otherwise I would have. I'm still planning to do an
exhaustive set of tests of different Ubuntu/Kubuntu/Lubuntu/etc releases in
the future, but that's no something I have time for now or in the near
future.

[Kevin (kevinshlee)]

I tried Lubuntu and got the same problem. It's only 2 hours. I tested it with Lubuntu 11.04 64bit. It was only a Live USB as all I got after installation was just a dark blank screen. Although I used only a Live USB, I upgraded the kernel.

I installed Ubuntu 11.10 again and can still see the same problem. What should I do now? Open a new bug ticket?

My laptop is Lenovo E520 and has an Intel i7 2620M CPU.
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/760131/comments/184

[Tim Gardner (timg-tpi)]

Herton - this patch would have come via stable had this kernel been in the support window, so I think its OK to leave it alone. It does not appear to be causing any regressions, and I've read reports that is has made a difference on some platforms. There are likely multiple reasons for increased power consumption, among which are mis-configured MSR on Intel, increased usage of compiz and 3D graphics, pulse audio, etc.

[Tim Gardner (timg-tpi)]

This patch improves the situation for at least one person (#168) so I'm marking this verification-done. Furthermore, this bug report has become unmanageable. We'll be looking at this issue during the LTS development cycle, so everyone please start your own bug with your hardware specifics attached.

[Jakbest (jakbest)]

I HAVE THE SAME PROBLEM, BUT MINE IS VERY WORST!!! MY BATTERY STARTS WITH 2:45 HOURS REMAINING (AND WITH A CORRECT ENERGY SAVER IT CAN BE MORE THAN 3 HOURS) AND AFTER EACH 10 MINUTES MY BATTERY HAS 30 MINUTES LESS.

CAN SOMEONE UNDERSTAND THAT THIS BUG IS A VERY IMPORTANT REGRESSION THAT LET ME TO NOT RECOMMEND GNU/LINUX TO MY FRIEND WHO HAVE A NOTEBOOK/NETBOOK? I CAN UNDERSTAND ALL BUT THIS BUG IS RELATED WITH THE FIRST ONE: IF UBUNTU WANTS TO BE BETTER THAN WINDOWS THIS BUG MUST BE RESOLVED, ELSE THERE IS NO CHANCE!! I DON'T KNOW IF IT IS VERY A PROBLEM FROM 2.6.38 TO 3.1 (AND PROBABLY MORE) BUT I KNOW THAT IS A PROBLEM OF 3 RELEASE OF UBUNTU!!!

I WOULDN'T BE ANGRY IF UBUNTU WAS DISGUSTING, BUT IT IS A PERFECT SYSTEM (WITH SOME PROBLEM OF COURSE) BUT I CAN'T UNDERSTAND WHY WE HAVE THAT REGRESSION

Sorry for my English, I'm AN ITALIAN USER OF UBUNTU (THIS TO UNDERSTAND THAT UBUNTU IS NOT A LITTLE AND NOT IMPORTANT OS)

WITH ANGER, JAKBEST

[florin (florin-arjocu)]

@jakbest
As far as I could see, it's mostly a kernel problem, so I guess making some pressure on kernel developers to manage a way around power management + better drivers.

[florin (florin-arjocu)]

[(continuing idea)..] might be a good idea.

[Kevin (kevinshlee)]

Based on the several comments above, this issue seems to be hardware-specific. For now, I've given up using Ubuntu but want to use it again as soon as possible so I'd like to create a bug ticket for my laptop. Could anyone please tell me what information I should include (preferably how I could as well)? Thank you.

[Seth Forshee (sforshee)]

On Mon, Oct 17, 2011 at 02:40:09PM -0000, Kevin wrote:
> Based on the several comments above, this issue seems to be hardware-
> specific. For now, I've given up using Ubuntu but want to use it again
> as soon as possible so I'd like to create a bug ticket for my laptop.
> Could anyone please tell me what information I should include
> (preferably how I could as well)? Thank you.

Kevin: The best way to create your bug is to open a terminal and run
'ubuntu-bug linux'. This will automatically attach the relevant
information form your machine to the bug when it is created. Thanks!

[Julian Kalinowski (julakali)]

Aside from the fact that this is unrelated to a power "regression":
Is the ubuntu/linux power consumption "bug" really device-specific?

On my Thinkpad X60s, ubuntu uses at least 4 Watts more as Windows does: roughly 12W instead of 8W (thats 50%) .
I haven't seen a single device since my good old T40, which actually consumes LESS OR EQUAL power when running linux.
Note the fact that Thinkpads are probably the best supported notebooks..

Now Intel claims to care about this (you all know powertop from http://www.lesswatts.org), but nothing seems to change.
I think, as some people already mentioned, as the ubuntu-community seems to care about mobile devices (it seems they do), they should care more about power consumption. Closing a bug without a solution doesn't do it for me.

Sorry for polluting this bug even more.

And btw, if anybody knows a (sub-)notebook (preferably featuring a trackpoint) with minimal power consumption on linux , please tell me.

[florin (florin-arjocu)]

Maybe this issue would need special attention from Apport application, an "if" in it's code, to be able to add multiple sets of attachments on the same bug (attachments from different people; as it is now, if the bug is there already you cannot automatically attach your logs). I think it will also be easier for developers to have it all in one place than having hundreds of separate bugs. This way we can also see the development of it and you bet I am interested in seeing if there are some changes, even for another hw configuration.

In the end, a bitter but true "joke": Nowadays most of the computers are mobile, so why should linux team care about "mobility"? Let's just wait until Google will invest to fix it for Android (or maybe it's a scheme to force paying private support). :)

[madbiologist (me-again)]

@Julian - There are actually 3 power regressions discussed in this bug. The biggest one seems to have occurred between the 2.6.37 kernel and Natty's 2.6.38 kernel, when Active-State Power Management (ASPM) was disabled. This was done as some BIOS's have broken support for ASPM and linux will hang on those machines when trying to use ASPM. It was therefore decided to play it safe and disable ASPM in linux for all hardware. This power regression is not very hardware specific at all - it affects any machine with a PCI Express motherboard.

You can force ASPM by adding pcie_aspm=force to your GRUB boot command line. This should work for many systems, but if your system's BIOS/hardware is in bad shape, you may get system hangs.

The fix in comment #164 (on the basis of which this bug was marked as fixed, in combination with comment #168) corrects a much smaller power regression in EPB (energy/performance bias), specfically the value of MSR_IA32_ENERGY_PERF_BIAS. This in fact may not even be a regression, but simply a new addition of support for an energy-saving feature. It is very hardware specific - MSR_IA32_ENERGY_PERF_BIAS was only introduced in the recent Intel "Westmere" Xeon CPUs and the even more recent Intel "Sandy Bridge" CPUs.

Older Intel hardware and non-Intel hardware does not use this MSR. You can also check whether your system supports this feature by looking for the "epb" flag in /proc/cpuinfo. Within the dmesg output should also be ENERGY_PERF_BIAS messages, if being used.

It is worth noting that the original reporter of this bug has an older CPU that does not support the energy performance bias MSR.

[madbiologist (me-again)]

@ Tim Gardner and Leann Ogasawara - further to comment #209, particularly the last sentence, I have reviewed this bug. The author of comment #168 has not tested the patch from comment #164 which you refer to in comment #201. They performed a "package-upgrade" to linux-image 2.6.38-11 (which version we are not sure) before the patch was even in natty-proposed. Comment #195 is vague, and comment #196 is about hardware which does not support the energy performance bias MSR.

Therefore, given the large number of subsequent comments stating that the patch has not reduced power consumption, and given the original reporter (I am led to believe they are a Canonical employee)'s hardware, can we really mark this bug as Fix Released, or even as Fix Committed?

However, marking as verification done for the SRU for the energy performance bias MSR seems appropriate - I agree with comment #200.

[Pako (elektrobank01)]

@ madbiologist

So now we have to upgrade our computers with the latest intel hardware?

[madbiologist (me-again)]

@ Kevin and anyone else with an Intel "Sandy Bridge" CPU who is experiencing even higher power consumption in Oneiric than in Natty - that is yet another power regression and should be reported in a new bug - see http://www.phoronix.com/scan.php?page=article&item=linux_31_power_regress&num=1

[Pako (elektrobank01)]

@ madbiologist:
Why Lubuntu last 4 and Ubuntu 1:45h. on my laptop? Same computer, same kernel.

[madbiologist (me-again)]

@ Pako - I wasn't suggesting that anyone should upgrade their computer to the latest Intel hardware. Sorry if it appeared that way. I was trying to explain that the so called "fix" for this bug is in fact limited to the latest Intel hardware. I think it is unfortunate and inappropriate that this bug has been marked as fixed when so many people (including the original reporter) do not and cannot benefit from the so called "fix".

[Pako (elektrobank01)]

madbiologist, Please answer my second question.

[madbiologist (me-again)]

@ Pako re comment #213 - I think you answered that question yourself in comment #197.

What is the output of lspci -vvnn on that computer? And what is the output of "glxinfo |grep render" (without the quotation marks) from Lubuntu and from Ubuntu? Please attach the information to this bug report (try to avoid pasting it into a comment if at all possible). Note that to run glxinfo you will need to first install the mesa-utils package.

[Pako (elektrobank01)]

Why would you need glxinfo? We talk about kernel power regression.

[Laurynas Biveinis (laurynas-biveinis)]

One more data point. On my Lenovo ThinkPad T410 (so, same as in comment #168) I was able to recover most of the Maverick-like battery life by updating to the newest BIOS. Before the upgrade powertop showed ~25W of battery drain w/o tuning, with pcie_aspm=force and all powertop suggestions done it went down to ~20W, and after the BIOS upgrade it is 16-17W, which is only slightly worse than with Maverick.

It can be further improved, now powertop shows the soundcard x2 taking 1.7W and used 100% (when no sound is played), that I will report separately.

[madbiologist (me-again)]

I was trying to answer your question. But you are correct, this bug is about kernel power regressions. Therefore your question in comment #213 is outside the scope of this bug. Please note that demanding an answer to a question, or a response to a comment, as you did in comment #215, is poor form. We are/were all trying to help each other here. Making such demands might also be against the Ubuntu code of conduct.

I have no idea why Ubuntu uses more power than Lubuntu on your laptop and I don't care any more.

[madbiologist (me-again)]

@ Laurynas - that is excellent news! I'm guessing that pcie_aspm=force now has no effect on power consumption?

[Laurynas Biveinis (laurynas-biveinis)]

@madbiologist. It still does. The system runs at 19-20W without it instead of 16-17W. Whatever the BIOS update fixed, it was not ASPM I think:

laurynas@laurynas-ThinkPad-T410:~$ dmesg | grep ASPM
[ 0.498093] ACPI FADT declares the system doesn't support PCIe ASPM, so disable it
[ 0.788492] ACPI _OSC control for PCIe not granted, disabling ASPM
[ 0.792706] ACPI _OSC control for PCIe not granted, disabling ASPM

[Nashvin (nashving)]

Hello everyone

I just want to find out if it is easy to observe this problem when using Live CD distributions. I want to try and test a few different versions of distributions (using Live CDs, so I don't mess up my current setup). Is there a difference in power consumption between a Live CD and an installed distro?

[reini (rrumberger)]

Well, for one Live CDs tend to run drom CD, which tends to use a bucketload of energy over time...
I'm not sure if you can compare Live CDs to other ones, but you definitely can't compare them to a system installed on a hard drive.

[madbiologist (me-again)]

@ Kevin and anyone else with an Intel "Sandy Bridge" CPU who is experiencing even higher power consumption in Oneiric than in Natty - further to comment #212, that issue is being tracked in bug #818830

[Carlos Moffat (carlos-eldiabloenlosdetalles)]

Hello,

I have a thinkpad x200 (sandy bridge), and I don't think my experience is very typical, but: I saw some of the same power regressions people are talking about, but I stuck with the latter kernels (I'm using 3.0.4 from the mainline site) because the older, energy-saving kernels had some trouble with the sandy bridge chip.

In any case, I was using Natty+Gnome Classic, but a few days I decided to take the plunge and update to Oneiric+Unity, and low and behold, I'm seeing *much better* power performance *with the same kernel I was using before*. With Natty, I could barely get power consumption down to maybe 13 W (no bluetooth or network, etc). As I write this, with bluetooth off but wifi on, the laptop is drawing ~9W, and I still have >6h to go with 65% battery. Suffering occasional unity freezes, so not everything is sunny, but...

One difference is that I set all the tunables using PowerTop, and the version that comes with Oneiric has more tunables than the previous one. Now, if I can only make them persistent over shutdowns and suspend/resume cycles...

In any case, this is just to say that in my case, not everything is about the kernel...

[tekstr1der (tekstr1der)]

@Carlos: Do you mean to say you've got a ThinkPad X220 (sandy bridge)? Or, as you stated, do you have an X200 (core 2 duo)?

This bug was originally reported regarding an X200s with core 2 duo, but was highjacked and subsequently "fixed" only for current Sandy Bridge processors. See comment #209 for a coherent explanation of the confusion in this bug report.

[Carlos Moffat (carlos-eldiabloenlosdetalles)]

@Tekstr1der: Yes, I meant a X220. But my point was that I saw a significant change in changing Ubuntu versions *without* changing the kernel. But maybe it's an edge case.

Cheers.

[Pako (elektrobank01)]

That's why Ubuntu should leave Xorg, urgently! Sooner, better. All of the power, performance, hangs, freeze..... will be fixed instantly when Wayland comes up.

[Launchpad Janitor (janitor)]

This bug was fixed in the package linux - 2.6.38-12.51

---------------
linux (2.6.38-12.51) natty-proposed; urgency=low

  [Herton R. Krzesinski]

  * Release Tracking Bug
    - LP: #860832

  [ Alex Bligh ]

  * SAUCE: net/netfilter/nf_conntrack_netlink.c: fix Oops on container
    destroy
    - LP: #843892

  [ Jesse Sung ]

  * SAUCE: Unregister input device only if it is registered
    - LP: #839238

  [ Leann Ogasawara ]

  * SAUCE: x86: reboot: Make Dell Latitude E6220 use reboot=pci
    - LP: #838402
  * SAUCE: x86: reboot: Make Dell Latitude E6520 use reboot=pci
    - LP: #833705

  [ Ming Lei ]

  * SAUCE: fireware: add NO_MSI quirks for o2micro controller
    - LP: #801719

  [ Stefan Bader ]

  * [Config] Include all filesystem modules for virtual
    - LP: #761809

  [ Tim Gardner ]

  * [Config] kernel preparation cannot be parallelized
  * [Config] Linearize module/abi checks
  * [Config] Linearize and simplify tree preparation rules
  * [Config] Build kernel image in parallel with modules
  * [Config] Set concurrency for kmake invocations
  * [Config] Improve install-arch-headers speed
  * [Config] Fix binary-perarch dependencies
  * [Config] Removed stamp-flavours target
  * [Config] Serialize binary indep targets
  * [Config] Use build stamp directly
  * [Config] Restore prepare-% target
  * [Config] Fix binary-% build target

  [ Upstream Kernel Changes ]

  * Revert "drm/i915: disable PCH ports if needed when disabling a CRTC"
    - LP: #814325, #838181
  * drm/i915: restore only the mode of this driver on lastclose (v2)
    - LP: #848687
  * cifs: fix possible memory corruption in CIFSFindNext, CVE-2011-3191
    - LP: #834135
    - CVE-2011-3191
  * befs: Validate length of long symbolic links, CVE-2011-2928
    - LP: #834124
    - CVE-2011-2928
  * gro: Only reset frag0 when skb can be pulled, CVE-2011-2723
    - LP: #844371
    - CVE-2011-2723
  * inet_diag: fix inet_diag_bc_audit(), CVE-2011-2213
    - LP: #838421
    - CVE-2011-2213
  * si4713-i2c: avoid potential buffer overflow on si4713, CVE-2011-2700
    - LP: #844370
    - CVE-2011-2700
  * Bluetooth: Prevent buffer overflow in l2cap config request,
    CVE-2011-2497
    - LP: #838423
    - CVE-2011-2497
  * crypto: Move md5_transform to lib/md5.c, CVE-2011-3188
    - LP: #834129
    - CVE-2011-3188
  * net: Compute protocol sequence numbers and fragment IDs using MD5,
    CVE-2011-3188
    - LP: #834129
    - CVE-2011-3188
  * x86, intel, power: Initialize MSR_IA32_ENERGY_PERF_BIAS
    - LP: #760131
  * x86, intel, power: Correct the MSR_IA32_ENERGY_PERF_BIAS message
    - LP: #760131
  * rt2x00: Serialize TX operations on a queue.
    - LP: #855239
  * ext4: Fix max file size and logical block counting of extent format
    file, CVE-2011-2695
    - LP: #819574
    - CVE-2011-2695
 -- Herton Ronaldo Krzesinski <email address hidden> Tue, 27 Sep 2011 16:19:57 -0300

[FrankSL (franksl)]

I have a HP 6710b laptop and still have this problem in Natty with latest updates. My system i a dual boot with windows xp, so I can easily compare battery life and with natty is way shorter than with windows.
Please tell me if I can provide any info.

[piccobello (piccobello)]

@Launchpad Janitor (janitor) wrote on 2011-10-25: #229

thanks a lot, however I did test the new kernel from natty-proposed, it does not solve the problem for me.
It does marginally improve over 2.6.38-11.50, but it is still a major regression compared
to 2.6.35-28.50. For me the problem is not only reduced battery life, but also increased
cpu temperature (easily above 70C) which makes it difficult to work, and of course is
not good for the cpu. I'm now running some extensive tests with powertop, discharging
the battery completely, which will take some time, I'll post the results here asap.

[madbiologist (me-again)]

@FrankSL - what CPU is in the HP 6710b?

@piccobello What CPU is in your PC?

[Pako (elektrobank01)]

Current way of rendering the GUI is dependent on CPU power, so battery on stronger processors runs longer, why, because the stronger CPU needs less time to render the GUI, it's just like in vehicles, as vehicle has more horsepower it consumes less fuel, because it takes less time to achieve the speed, and because GPU is 10 times stronger than CPU, battery on Windows lasts more. In Linux we hardly have Open GL 1.5. support. So, I think even in this stage Wayland should be the default server on Ubuntu because of this.

[Kevin (kevinshlee)]

Thank you Seth Forshee and madbiologist. As you suggested I created a new bug ticket. bug #883403

[piccobello (piccobello)]

After some more extensive testing I cannot confirm the regression anymore on my Panasonic CF-T5:
forcing pcie_aspm and manually setting powersave:

echo powersave > /sys/module/pcie_aspm/parameters/policy

solves the problem for me as well.
From /proc/cpuinfo:
vendor_id : GenuineIntel
cpu family : 6
model : 14
model name : Genuine Intel(R) CPU U1400 @ 1.20GHz

I made a script for pm-powersave, attached as pcie_aspm, and put it in /etc/pm/power.d

I made a table to summarize the results of the tests, reporting powertop results on 4 hours
Wakeups/s (lower=better), C3% (perc. of time in higest C state, higher=better), PLow% (perc. of time
at lowest CPU freq, higher=better), battery duration (hh:mm, based on battery-stats log), avg temperatures
of the two available sensors (estimate based on collectd graphs). All tests were run with screen brightness
10/21, dpms on (switching off the screen after 10min), wifi off (in normal conditions the battery lasts less then
half the times obtained in the tests, but it's still pretty good).

Kernel | Wakeups/s | C3% | PLow% | Power | Battery | Temp0 C | Temp1 C |
2.6.35-28.50 +aspm 50.7 93.6 77.0 5.8W 15:15 38 39
2.6.38-11.50 +aspm 53.7 94.4 81.8 4.0W 16:23 37 38
2.6.38-12.51 +aspm a 47.2 94.1 84.6 8.7W 8:57 38 42
2.6.38-12.51 +aspm b 46.1 93.2 70.2 4.8W 14:52 38 39

The thing I realized running the tests is that acpi estimate of battery state (and power) is not so reliable,
therefore I had very different results with when starting the tests as soon as the charge was 100%, and
when it was left powered for a few hours while at 100%. That was the only difference among the two runs
with 2.6.38-12.51, marked a and b in the table. As you can see, repeating the test after some time with the
ac adapter plugged in gave me results which are comparable to maverick.

Thanks a lot to all contributors to this very long bug report!

[Jaxon (bjackson0971)]

There's a tweet today from @michaellarabel about a kernel patch to fix ASPM. It looks like this one:

[PATCH] pci: Rework ASPM disable code
https://lkml.org/lkml/2011/11/10/467

It applies to 3.1 source with an offset warning but no errors. I applies cleanly to 3.2-rc1. I haven't tested bootup yet with the patch applied.

[michaellarabel (michael-michaellarabel)]

Yes it's that patch that's fixed the "out of the box" ASPM support on all of the hardware I've hit this afternoon - http://www.phoronix.com/scan.php?page=article&item=linux_aspm_solution&num=1

[Jaxon (bjackson0971)]

There were some more patches posted that disable ASPM on specific hardware, like Microsoft does.

https://lkml.org/lkml/2011/11/11/178

[carlosv (cvedovatti)]

Matthew Garrett proposed a patch for the ASPM Linux problem:

https://lkml.org/lkml/2011/11/10/467

[oussama (obounaim)]

Stilling facing this bug with sony vaio ubuntu 10.04.3 2.6.32-38-generic
when i m using ubuntu the battery drop down in half time than windows

[madbiologist (me-again)]

Oussama - This bug is about a power consumption regression between Maverick and Natty. If you are using Lucid (10.04.x) then you have a different bug.

The patch/es described in comments 236-239 will be available in the upstream kernel 3.3. Canonical has back-ported this patch into the forthcoming Ubuntu 12.04 "Precise Pangolin"'s 3.2-based kernel. Ubuntu 12.04 alpha 2 was just released today.

[madbiologist (me-again)]

The patch I mentioned in comment #241 is now also available in the upstream 3.2.5 kernel, and more importantly for Oneiric, in the upstream 3.0.20 kernel. PPA's of these kernels are available at http://kernel.ubuntu.com/~kernel-ppa/mainline/

[madbiologist (me-again)]

The 3.2.0-15.24 kernel in Ubuntu 12.04 "Precise Pangolin" repository has been rebased to upstream 3.2.5.

[Lorenzo Bettini (bettini)]

my laptop does not seem to support ASPM:

dmesg | grep ASP
[ 0.963968] ACPI FADT declares the system doesn't support PCIe ASPM, so disable it
[ 1.328171] ACPI _OSC control for PCIe not granted, disabling ASPM

but it looks like it supports epb:

dmesg | grep ENE
[ 0.000241] ENERGY_PERF_BIAS: Set to 'normal', was 'performance'

so what could I do to reduce power consumption? with Linux battery lasts less than 2 hours, in windows lasts almost 5 hours!