cpufreq locked in slowest speed

[Robert Collins]

Binary package hint: linux-source-2.6.20

I have a dell X1 laptop, and with the 2.6.20-9 kernel, I sometimes see it get 'locked' into the slowest cpu freq that it understands.

The governor is set to 'performance'.

I *think* that this only occurs after I have done a suspend to ram and resume cycle.

[Kyle McMartin (kyle)]

Can you paste the output from the following files in /sys/devices/system/cpu/cpu0/cpufreq/

scaling_driver
scaling_governor
scaling_(min|max)_freq
scaling_setspeed
scaling_available_governors
scaling_available_frequencies

Thanks!
  Kyle

[Robert Collins (lifeless)]

On Wed, 2007-03-07 at 19:10 +0000, Kyle McMartin wrote:
> Can you paste the output from the following files in
> /sys/devices/system/cpu/cpu0/cpufreq/

> scaling_setspeed
This doesn't' exist. The rest are below.

Right this instant its working correctly, but it was locked in slow
again last night - it seems to come and go.

> Info

file scaling_driver
acpi-cpufreq
file scaling_governor
performance
file scaling_min_freq
600000
file scaling_max_freq
1100000
file scaling_available_governors
userspace powersave ondemand conservative performance
file scaling_available_frequencies
1100000 800000 600000

--
GPG key available at: <http://www.robertcollins.net/keys.txt>.

[Robert Collins (lifeless)]

its playing up right now: I have power attached FWIW.
scaling_setspeed is still non-existant

for fname in scaling_driver scaling_governor scaling_min_freq scaling_max_freq scaling_available_governors scaling_available_frequencies; do echo file $fname; cat $fname; done
file scaling_driver
acpi-cpufreq
file scaling_governor
performance
file scaling_min_freq
600000
file scaling_max_freq
600000
file scaling_available_governors
userspace powersave ondemand conservative performance
file scaling_available_frequencies
1100000 800000 600000

[Robert Collins (lifeless)]

foo

[Robert Collins (lifeless)]

I've done some testing, it seems to be broken only if I have booted up from cold without power. If I boot with power it its all good. If I hibernate when its broken, plugin power in and resume, then its broken until I unplug power, at which point it comes good.

[Robert Collins (lifeless)]

more data: overnight, without me suspending or hibernating, its gotten locked into slow-mode.

[Gabriel Ambuehl (gabriel-ambuehl)]

This seems to happeon on my Thinkpad T60 (Core 2 Duo 2 Ghz) too:

sudo cpufreq-info
cpufrequtils 002: cpufreq-info (C) Dominik Brodowski 2004-2006
Report errors and bugs to <email address hidden>, please.
analyzing CPU 0:
  driver: acpi-cpufreq
  CPUs which need to switch frequency at the same time: 0 1
  hardware limits: 1000 MHz - 2.00 GHz
  available frequency steps: 2.00 GHz, 1.67 GHz, 1.33 GHz, 1000 MHz
  available cpufreq governors: userspace, ondemand, powersave, conservative, performance
  current policy: frequency should be within 1000 MHz and 1000 MHz.
                  The governor "performance" may decide which speed to use
                  within this range.
  current CPU frequency is 1000 MHz (asserted by call to hardware).
analyzing CPU 1:
  driver: acpi-cpufreq
  CPUs which need to switch frequency at the same time: 0 1
  hardware limits: 1000 MHz - 2.00 GHz
  available frequency steps: 2.00 GHz, 1.67 GHz, 1.33 GHz, 1000 MHz
  available cpufreq governors: userspace, ondemand, powersave, conservative, performance
  current policy: frequency should be within 1000 MHz and 1000 MHz.
                  The governor "performance" may decide which speed to use
                  within this range.
  current CPU frequency is 1000 MHz (asserted by call to hardware).

So while Linux knows there are several modes, it only allows cpufreq to use the slowest of them...

[Gabriel Ambuehl (gabriel-ambuehl)]

After a warm reboot (and using Windows Vista in between), I now have the full
range again:

  current policy: frequency should be within 1000 MHz and 2.00 GHz.

[José P Valdés (sevmpe-deactivatedaccount)]

I have the same problem in a Dell 640m with a Core 2 Duo 2GHz. After resuming from suspend, one of the cores is locked at 1GHz, always the one labeled CPU1. I have tried to use cpufreq-selector to force other governors (it is in ondemand) but then something weird happens: when I use it either with -c 0 or -c 1 it always changes the governor of CPU0.

[Nic (ntetreau)]

Same problem as Jose on a Sony Vaio VGN-SZ3XWPC, this is the same processor as Jose I believe.

[Gabriel Ambuehl (gabriel-ambuehl)]

I think I just managed to trigger this without suspending etc.

I cold-booting, had the full spectrum of frequencies, decided to run the tpfancontrol script (which overrides the inbuilt fan control logic) and the cpufreq spectrum got locked between 1Ghz and 1Ghz:

  current policy: frequency should be within 1000 MHz and 1000 MHz.

Running cpufreq-set doesn't change anything.

[Gabriel Ambuehl (gabriel-ambuehl)]

I just tried linux 2.6.21.1 (as from kernel.org, no custom patches) on my Thinkpad T60 and after a short suspend to RAM, my CPU still is able to scale!

However, when you build this kernel, you won't have Ubuntu's restricted modules so the ipw3945 won't work out of the box, for example.

[Gabriel Ambuehl (gabriel-ambuehl)]

This issue persists on 2.6.20-16, even without any suspends in between, my frequency just got locked to 1Ghz on my Thinkpad T60.

[Gabriel Ambuehl (gabriel-ambuehl)]

I stand corrected, it just happened to me on a 2.6.21.1 kernel, too.

[Loic Dachary (dachary)]

Same problem here on a 2.6.18-vserver on a Dell 9400 dual core.

[siripiscuipi (jzuniga)]

I have the same problem here on kernel 2.6.20-16-generic.

/proc/cpuinfo

processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 13
model name : Intel(R) Pentium(R) M processor 1.73GHz
stepping : 8
cpu MHz : 798.000
cache size : 2048 KB
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 2
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat clflush dts acpi mmx fxsr sse sse2 ss tm pbe nx up est tm2
bogomips : 1597.42
clflush size : 64

[MarkBrier (mark-brier)]

This occurs on my Thinkpad x31 Pentium-M 1.6GHz

After a hibernate resume cycle, the processor will not go above 1GHz. The option to go above 1GHz is still shown but doesn't work. After a reboot or a cold boot everything works as it should (i.e. upto 1.6GHz)

mark@thinkpad:~$ cat /proc/cpuinfo
processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 9
model name : Intel(R) Pentium(R) M processor 1600MHz
stepping : 5
cpu MHz : 1000.000
[snip]...

mark@thinkpad:~$ cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_frequencies
600000 800000 1000000 1200000 1400000 1600000

mark@thinkpad:~$ cpufreq-info
cpufrequtils 002: cpufreq-info (C) Dominik Brodowski 2004-2006
Report errors and bugs to <email address hidden>, please.
analyzing CPU 0:
  driver: centrino
  CPUs which need to switch frequency at the same time: 0
  hardware limits: 600 MHz - 1.60 GHz
  available frequency steps: 600 MHz, 800 MHz, 1000 MHz, 1.20 GHz, 1.40 GHz, 1.60 GHz
  available cpufreq governors: conservative, userspace, ondemand, powersave, performance
  current policy: frequency should be within 600 MHz and 1000 MHz.
                  The governor "performance" may decide which speed to use
                  within this range.
  current CPU frequency is 1000 MHz.

mark@thinkpad:~$ sudo cpufreq-set -f 1600000
mark@thinkpad:~$ cpufreq-info -f
1000000

mark@thinkpad:~$ uname -r
2.6.20-16-generic

[MarkBrier (mark-brier)]

Just to clarify, I can reduce the speed of the CPU, I cannot however increase above 1GHz

[Bram Stolk (b-stolk)]

I have the same, using

I noticed a mis-match between scaling_max_freq and cpuinfo_max_freq
Even when switching to perfomance governor, it is stuck at 800MHz

root@mediacenter:/sys/devices/system/cpu/cpu0/cpufreq# cat cpuinfo_max_freq
1600000
root@mediacenter:/sys/devices/system/cpu/cpu0/cpufreq# cat scaling_max_freq
800000
root@mediacenter:/sys/devices/system/cpu/cpu0/cpufreq# uname -a
Linux mediacenter 2.6.22-14-generic #1 SMP Sun Oct 14 23:05:12 GMT 2007 i686 GNU/Linux

  Bram

[Bram Stolk (b-stolk)]

Why is this bug still set to 'incomplete'?
There is lots of data in this report.
If you need more, please specify.

[lunomad (damon-metapaso)]

I too have the same system and same bug as the original post: Dell Latitude X1 notebook.
This problem began for me when I upgraded from Feisty to Gutsy.

My kernel is: 2.6.22-14-generic

I scanned through Related Bugs:
Bug #93404
Bug #67341
Bug #132271
all have similar problems, but can't find exactly my situation as for me the problem is much more sporadic. Sometimes I relaunch powernowd and the scaling works for awhile up to the system max of 1100MHz, then after awhile I only have 600MHz and 800MHz, then it just locks onto 600MHz. As far as I can tell, without reboot or restart of powernowd, the issue does not self-resolve. Although, as I type this, I see the CPUfreq applet suddenly display 800MHz. I still cannot get to 100%.

Please note that the X1 has no fan and CPU temperatures run high. Idle in a cold room is 52C, 1100MHz in this same room can get as high as 70C.

[lunomad (damon-metapaso)]

(Follow-up from previous comment)

I've done a closer study of my Dell X1 and as far as the Dell X1 is concerned (and may be true for others as well), this issue is definitely related to cpu heat. The X1 is an underpowered (1100MHz) Pentium M with purely passive cooling (no fan), so heat is a critical element.

Something in Linux is setting my heat ceiling at 65C. When the computer has a lot of work to do, powernowd sets the clock to full speed (1.1GHz) and the temperature slowly creeps up to 65C. When it reaches 65, the computer throttles down to 800MHz. This speed value is written into /sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq, and so powernowd has only 600MHz and 800MHz to choose between.

Even so, at 800MHz the computer will still creep up to 65C, and so when it reaches this temp, it will throttle down to 600MHz only.

The problem is that normal operation doesn't return to the system until the CPU temp falls below the lower bound of 50C. That is, I can't get 800MHz or 1100MHz speeds until the processor is much cooler. Back in my EE days, we called this difference bewteen activation and disactivation a "hysteresis". A delta-Temp (or hysteresis) of 15C on this system is too high.

So. I have identified two problems:
1. Thermal Threshold is set too low (this computer should run up to 70C (as it used to do under Feisty, kernel 2.6.16).
2. Hysteresis (that is, Delta-Temp) is set too wide...should be reduced to 5C or 7C, so that the computer can throttle up and throttle down.

I don't know how to set/change any of these values. Any help out there?

Thanks.

[Robert Collins (lifeless)]

On Sun, 2007-11-25 at 21:39 +0000, lunomad wrote:
>
>
> So. I have identified two problems:
> 1. Thermal Threshold is set too low (this computer should run up to
> 70C (as it used to do under Feisty, kernel 2.6.16).
> 2. Hysteresis (that is, Delta-Temp) is set too wide...should be
> reduced to 5C or 7C, so that the computer can throttle up and throttle
> down.
>
> I don't know how to set/change any of these values. Any help out
> there?

Excellent analysis - thanks!

I'm not sure how to set them either; but I imagine the kernel module
source for scaling will be relevant.

-Rob
--
GPG key available at: <http://www.robertcollins.net/keys.txt>.

[lunomad (damon-metapaso)]

Rob:

I finally got a chance to do some hunting on the internet for a solution to this problem. The problem appears to be rooted in unsettable BIOS thermal tables. Downgrading to BIOS A02 appears to solve the problem, but as I run dual-screen I don't think I'll be doing this (the A02 has earlier Intel915 graphics firmware). I might try a kernel downgrade to see if the problem can be resolved with my previous kernel (I don't recall this same issue, but perhaps I just didn't notice). I used to get an error 7@1000:10000 on boot, which has disappeared. Perhaps something in there?

A whole bunch of great commentary here:

http://eightflat.org/tracesofmartinsstate/2005/06/03/linux-on-dell-x1/#comment-13235

There are two good Linux/Dell X1 sites:

http://gentoo-wiki.com/Talk:HARDWARE_Dell_Latitude_X1#CPU_down-throttling (gentoo)
http://eightflat.org/tracesofmartinsstate/ (Fedora)

[lunomad (damon-metapaso)]

Kyle and Rob:

I think this bug should be marked closed, as it appears to be a BIOS-dependent setting particular to the Dell X1 as per the original post.

Other posters on this thread reporting similar behavior seem to be duplicates of other cpufreq bugs...not sure what the bug policy on this is, but as far as the X1 is concerned, I don't believe anyone can help us except for Dell's BIOS programmers.

Thanks,
Damon

[Ev Kontsevoy (biz-kontsevoy)]

NO! This bug MUST NOT be closed, because it has nothing to do with Dell. Bunch of people including myself have reported issues with Lenovo laptops as well. And no, it should not be related to BIOS since a simple reboot fixes the problem.

[Gabriel Ambuehl (gabriel-ambuehl)]

I agree, I have seen this on a Lenovo T60 myself. And it also occurs on my
Latitude D830. For the most part, I can fix it by removing acpi_cpufreq and
reloading it, but at times the module is in use and thus cannot be removed.

I think this is a kernel bug of some sort, as it occurs in vanilla kernel.org
kernels as well.

[Robert Collins (lifeless)]

On Wed, 2007-12-26 at 09:58 +0000, Gabriel Ambuehl wrote:
> I agree, I have seen this on a Lenovo T60 myself. And it also occurs on my
> Latitude D830. For the most part, I can fix it by removing acpi_cpufreq and
> reloading it, but at times the module is in use and thus cannot be removed.
>
> I think this is a kernel bug of some sort, as it occurs in vanilla kernel.org
> kernels as well.

It is extremely likely that you have an bug with the *same symptoms*. If
the fault lies in the DSDT then it is:
a) not a kernel bug
b) specific to the hardware

And thus its appropriate for you to report *separate* bugs for your
separate DSDT tables - it is hardware specific.

Cheers,
Rob

--
GPG key available at: <http://www.robertcollins.net/keys.txt>.

[Launchpad Janitor (janitor)]

Beginning with the Hardy Heron 8.04 development cycle, all open Ubuntu kernel bugs need to be reported against the "linux" kernel package. We are automatically migrating this bug to the new "linux" package. However, development has already began for the upcoming Intrepid Ibex 8.10 release. It would be helpful if you could test the upcoming release and verify if this is still an issue - http://www.ubuntu.com/testing . If the issue still exists, please update this report by changing the Status of the "linux" task from "Incomplete" to "New". We appreciate your patience and understanding as we make this transition. Thanks!

[Leann Ogasawara (leannogasawara)]

The Ubuntu Kernel Team is planning to move to the 2.6.27 kernel for the upcoming Intrepid Ibex 8.10 release. As a result, the kernel team would appreciate it if you could please test this newer 2.6.27 Ubuntu kernel. There are one of two ways you should be able to test:

1) If you are comfortable installing packages on your own, the linux-image-2.6.27-* package is currently available for you to install and test.

--or--

2) The upcoming Alpha5 for Intrepid Ibex 8.10 will contain this newer 2.6.27 Ubuntu kernel. Alpha5 is set to be released Thursday Sept 4. Please watch http://www.ubuntu.com/testing for Alpha5 to be announced. You should then be able to test via a LiveCD.

Please let us know immediately if this newer 2.6.27 kernel resolves the bug reported here or if the issue remains. More importantly, please open a new bug report for each new bug/regression introduced by the 2.6.27 kernel and tag the bug report with 'linux-2.6.27'. Also, please specifically note if the issue does or does not appear in the 2.6.26 kernel. Thanks again, we really appreicate your help and feedback.

[Ciaran Liedeman (ciaran-liedeman)]

Latest version of Intrepid Ibex, kernel version 2.6.27-7. same "bug" reported - chipset Intel 965PM+ICH-M, T7500 2.20Ghz Intel Core 2 duo processor capped within 2 minutes of start up on warm boot. Processor capped to 800Mhz, governors can be changed but this has no effect on CPU clock speed.

Must still try cold boot but based to what I've read here this will only solve ( alleviate.. ) the problem temporarily,
 also my laptop is a Fujitsu Siemens amilo Xi 2428.

I only discovered this problem today and this may be a stupid question but does the CPU frequency monitor create this problem or uncover it - having my processing power less than halved constantly is far from ideal.

I will post the results of a cold boot tomorrow.

[Gabriel Ambuehl (gabriel-ambuehl)]

Apathy wrote:
> I only discovered this problem today and this may be a stupid question
> but does the CPU frequency monitor create this problem or uncover it -
> having my processing power less than halved constantly is far from
> ideal.
>
> I will post the results of a cold boot tomorrow.
>
>
On my Latitude D830 it does not seem to occur anymore in 8.10, not even
after resuming from suspend.

[Ciaran Liedeman (ciaran-liedeman)]

I tried cold boot this morning and the CPU speeds were immediately locked, the cpuinfo_max_freq does not match the scaling_max_frequency.

I don't really every move my laptop around much so being able to control the cpu speed is not really important to me - I'll rather investigate disabling the governor - if that's possible

[Gabriel Ambuehl (gabriel-ambuehl)]

Apathy wrote:
> I tried cold boot this morning and the CPU speeds were immediately
> locked, the cpuinfo_max_freq does not match the scaling_max_frequency.
>
> I don't really every move my laptop around much so being able to control
> the cpu speed is not really important to me - I'll rather investigate
> disabling the governor - if that's possible
>
>
What happens if you simply set it to the performance governor upon boot
(or possibly even at kernel compile time)? That should have it running
at full speed all the time.

[Ciaran Liedeman (ciaran-liedeman)]

Do you mean a command for instance
$ sudo cpufreq-selector -f 2201000
because this command seems to have no effect.

By boot I assume you mean before I log and I have no idea how to do that, I have never compiled the kernel myself before

[Ciaran Liedeman (ciaran-liedeman)]

I solved the problem, it was heat related - when my laptop is in quiet mode the clock speed gets capped. There is a longer explanation for this but its not relevant :) Bug averted?

[James Ward (jamesward)]

This is happening on my Lenovo W500. After some time (potentially heat related) the cpufreq drops to 800 Mhz and just gets stuck there. Changing the frequency and the governor have no effect.

[Andy Whitcroft (apw)]

It has been a long time since anyone commented on this bug. This might indicate the issue is no longer present on releases you are running. Can you confirm this issue exists with the most recent Jaunty Jackalope 9.04 release - http://www.ubuntu.com/news/ubuntu-9.04-desktop . Please let us know your results. Thanks.

[A. Carboni (acarboni)]

I confirm this bug in Jaunty 9.04. I had the same with Hardy 8.04 but the problem was less frequent. I have a latitude d830 since march 2008. That time, I used a debian and all was ok. I installed 8.04 after some months and all was ok. Every while and then, I upgraded the bios up to release A14 and I suspect that the upgrade could be the cause. The processors not only get locked to 800mhz but the scaling is so strong that the system is pretty unusable. When the problem arises, it vanishes after some time (minutes, usually less than 1 hour). The problem seems related to the temperature. I tried even kernel 2.6.30.rc4 with the same results, with/without the nvidia driver.

Thanks for the help,
AC

[C. Brayton (cbrayton-boizebueditorial)]

Problem persists in Lucid 10.04Alpha1 (2.6.312-10) as in all releases after acpi-cpufreq was compiled in rather than left as a module. I cannot detect what is triggering the cpufreq governor to suddenly be set to that "between minimum and minimum," according to cpufreq-info. That it happens after a period of time in operation suggests it could be a temperature threshhold, but I thought I reset mine to 85-100 (reasonable for my Latitude D620, which tends to run between 60-75 degrees C.)

I am not seeing any kernel messages to this effect, though.

I am a poetry-major lower-end poweruser, and this has been driving me nuts for weeks and weeks. On suggestion of various fora, I recompiled a buggy DSDT, played with commandline options at bootup, and other stuff. Tried powernowd -d -vvv and noticed that it seemed to be working at first, resetting freqs according to the rules, at some point its started showing close to 100% CPU usage and indicating freq resets to the maximum that were not actually occurring.

Funny, though: it seems as though the Lucid daily LiveCD, booted off a USB stick, does not have this problem.

[Bharath Krishnan (bharath-krishnan-deactivatedaccount)]

I see this on my dell latitude e6400 after I upgraded to lucid. Didn't have this problem in karmic. Once for some reason if then fan comes on, the cpu max freq is stuck at 800Mhz and the fan doesn't switch off. Lucid is very frustrating for an LTS release, a bunch of things that were working in karmic are now broken (I know it isn't helpful to say that here, but ... ). Proxy handling: broken, External monitors: broken, Function keys: broken, sbackup: broken, list goes on.

[Alex Tomic (atomic777)]

I've commented on a similar bug to this for the Thinkpad T43 (bug#519142), but it seems this is a more general problem.

After some more debugging I've also discovered that my throttling issue is heat-related. For whatever reason, it seems my max cpu speed gets locked at 800mhz when my CPU goes above 53C and does not get reset to 2ghz until the temperature drops to around 51C. I have a 'watch cpu-info' running and I can see the policy change back and forth, so I know something is doing it.

My main problem is that I cannot figure out _what_ is setting this 53C temperature limit. Since I never had this problem on Karmic I know it is not BIOS-related since I did not have this problem on earlier versions.

[Jeremy Foshee (jeremyfoshee)]

This bug report was marked as Incomplete and has not had any updated comments for quite some time. As a result this bug is being closed. Please reopen if this is still an issue in the current Ubuntu release http://www.ubuntu.com/getubuntu/download . Also, please be sure to provide any requested information that may have been missing. To reopen the bug, click on the current status under the Status column and change the status back to "New". Thanks.

[This is an automated message. Apologies if it has reached you inappropriately; please just reply to this message indicating so.]

[ilpirata (gdistasi)]

I still have this problem on Kubuntu 10.10.

After a while the clock gets stuck to 800Mhz; no way to change it. Sometimes it recovers for a while, then after a while it gets again stuck at 800 Mhz (the minimum).

My pc is a Dell XPS 1330.

Regards,
Giovanni

[Pete (pvandoren)]

Re-opening since I have the same issues with 10.04 LTS 2.6.32-26-generic #48-Ubuntu SMP Wed Nov 24 10:14:11 UTC 2010 x86_64 GNU/Linux

and other users have stated they have it with 10.10. cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_frequencies
2535000 2534000 1600000
cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq
1600000

Current CPU temp is showign 51C.. Which is not unreasonable for a plugged in laptop. But I do believe it is heat related as far as when this issue kicks in.

Pete
and scaling_max_frequency shows:

I have a dell precision M6400 with a quad core processor, and scaling_available_frequencies shows:

[Pete (pvandoren)]

Sorry: Last post got scrambled somehow, but I think you can tell what's going on.

I just noticed some other interesting information: It doesn't effect all CPU cores at the same point. Currently my system is in a state where cpufreq/scaling_max_freq = 1.6 GHz for CPU 0 and CPU1, but CPU2 and CPU3 show 2.535 GHz.

cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq
1600000
cat /sys/devices/system/cpu/cpu1/cpufreq/scaling_max_freq
1600000
cat /sys/devices/system/cpu/cpu2/cpufreq/scaling_max_freq
2535000
cat /sys/devices/system/cpu/cpu3/cpufreq/scaling_max_freq
2535000

The max temp I saw was around 63C.

Pete

[Jeremy Foshee (jeremyfoshee)]

Pete,
   Please open a new bug for your issue. Closing this bug as it originated in 2007.

~JFo