'ondemand' failing under Feisty on Turion64 (Acer Ferrari 4005)

Bug #80813 reported by Dan Lenski
6
Affects Status Importance Assigned to Milestone
powernowd (Ubuntu)
Invalid
Undecided
Unassigned
Nominated for Feisty by Stefano Rivera

Bug Description

Binary package hint: powernowd

I'm running Feisty on an Athlon 64 (socket 939) system.

$ uname -a
Linux tonquil 2.6.19-7-generic #2 SMP Mon Dec 4 12:39:22 UTC 2006 x86_64 GNU/Linux

When the computer boots, the powernow-k8 module is loaded:

$dmesg | grep powernow
[ 93.306823] powernow-k8: Found 1 AMD Athlon(tm) 64 Processor 3500+ processors
 (version 2.00.00)
[ 93.306864] powernow-k8: 0 : fid 0xe (2200 MHz), vid 0x6
[ 93.306866] powernow-k8: 1 : fid 0xc (2000 MHz), vid 0x8
[ 93.306868] powernow-k8: 2 : fid 0xa (1800 MHz), vid 0xa
[ 93.306871] powernow-k8: 3 : fid 0x2 (1000 MHz), vid 0x12

However, the powernowd process never starts up. I can start it manually by running /etc/init.d/powernowd start, in which case it runs fine.

Under Edgy, I never had this problem. Powernowd would start up reliably every time I booted.

Revision history for this message
Stefano Rivera (stefanor) wrote :

This seems to be by design. Instead of starting, it tells the kernel to use the ondemand governer.

However, it doesn't seem to do this correctly, and my Turion64 (Acer Ferrari 4005) just sits at the lowest speed (800MHz), constantly.

Changed in powernowd:
status: Unconfirmed → Confirmed
Revision history for this message
Paul Sladen (sladen) wrote :

stefano: what happens if you use something CPU intensive, does the speed then ramp up after a while? Can you paste the output of the following:

  grep . /sys/devices/system/cpu/cpu*/cpufreq/scaling_[dg]*

Revision history for this message
Paul Sladen (sladen) wrote :

As noted, the CPU scaling is now done where-ever possible with the kernel CPU scaling governors, rather than userspace.

Revision history for this message
Stefano Rivera (stefanor) wrote :

Yes the speed ramps up after a *while* - i.e. at least a minute. It ramps up to 1.8GHz (90% of full speed).

$ grep . /sys/devices/system/cpu/cpu*/cpufreq/scaling_[dg]*
/sys/devices/system/cpu/cpu0/cpufreq/scaling_driver:powernow-k8
/sys/devices/system/cpu/cpu0/cpufreq/scaling_governor:conservative

I don't think conservative has the right behaviour for plugged-in operation. When I'm plugged in, I want my processor to ramp up *immediatly*. This behaviour is more suited to batter operation, and even then, it will disappoint people...

Revision history for this message
Paul Sladen (sladen) wrote :

Yeah, but there's a different issue here; '/etc/init.d/powernowd' should be selecting 'ondemand' by default during boot.

Did you say you'd attempted to installed/uninstall the 'powernowd' package at some point. Could you check you're using the default configuration for 'powernowd' and that the package is still correctly installed?

Revision history for this message
Stefano Rivera (stefanor) wrote :

OK, you're right. I just re-installed powernowd, and it does choose ondemand when it's initscript is started, however *something* else is changing the governer to powersave when I go off AC, and conservative when I plug in again.

If the conservative governor worked well on my machine, of course this wouldn't be a problem...

Revision history for this message
Stefano Rivera (stefanor) wrote :

Hmm, it seems to have been fixed for me, either in ubuntu, or by http://foodfight.org/log/Hardware/laptop-running-hot.html

Changed in powernowd:
status: Confirmed → Rejected
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.