cpufreq-applet doesn't do scaling

[Hannes Ovrén]

When I run kernel 2.6.15-20-k7 on my AMD Turion64 the gnome cpufreq-applet doesn't handle frequency scaling at all.

When I ran the 2.6.15-19-386 kernel it did work very well, but I can't run that kernel now since I updated my nvidia drivers which were incompatible with the -19 kernel.

#> dmesg | grep powernow
[4294697.511000] powernow-k8: Found 1 AMD Athlon 64 / Opteron processors (version 1.50.4)
[4294697.514000] powernow-k8: 0 : fid 0x0 (800 MHz), vid 0x1a (900 mV)
[4294697.514000] powernow-k8: 1 : fid 0x8 (1600 MHz), vid 0xe (1200 mV)

Powernowd doesn't give me any messages either.

So now my processor is constantly running at 1600 MHz instead of scaling between 800 and 1600 as it did before. It has the effect that my computer gets warmer than necessary.

[Brian Brunswick (brian-ithil)]

I had something like this, but it was due to something in xorg being cpu hungry and the usage never dropping low enough.

It would be worth seeing if things like
echo 800000 > /sys/devices/system/cpu/cpu0/cpufreq/scaling_setspeed
work, or adding
OPTIONS="-q -l50"
to /etc/default/powernowd to set the lower cpu limit to 50%. If that makes it work, then the problem is not the kernel scaling driver, but the cpu hungry xorg.

[Hannes Ovrén (kigurai)]

I found the problem. It was nothing wrong with the scaling. It was a
process eating all my CPU.
Guess we can close this bug.

2006/5/15, Brian Brunswick <email address hidden>:
> I had something like this, but it was due to something in xorg being cpu
> hungry and the usage never dropping low enough.
>
> It would be worth seeing if things like
> echo 800000 > /sys/devices/system/cpu/cpu0/cpufreq/scaling_setspeed
> work, or adding
> OPTIONS="-q -l50"
> to /etc/default/powernowd to set the lower cpu limit to 50%. If that makes it work, then the problem is not the kernel scaling driver, but the cpu hungry xorg.
>
> --
> cpufreq-applet doesn't do scaling
> https://launchpad.net/bugs/39010
>

[Adam Petaccia (mighmos)]

Rejecting based on previous comment.