ATI x1600 overheating on Ubuntu 10.10 64bit, laptop

[Porter]

I have a friend running vanilla Ubuntu 10.10 x64 (desktop) on a Compaq nx9420 laptop, which has an ATI x1600 integrated.

He is getting severe, rapid overheating of the GPU, to over 110 C and shutdown.

He disassembled the unit, checked the heat pipes and fan etc, and reassembled, same problem.

I had him disable KMS in Grub and reboot, didn't fix it. Verified that KMS is disabled in Xorg.0.log, but even at the GDM login or in textmode using Ctrl+Alt+F1 he is getting temperature rapidly climbing over 100 C.

This behavior does not occur in Windows on the same laptop, GPU stays cool and happy.

I'm stumped... any ideas folks?

Thanks,
Porter

[darkbasic]

mount -t debugfs none /sys/kernel/debug
echo profile > /sys/class/drm/card-0/device/power_method
echo low > /sys/class/drm/card-0/device/power_profile

P.S.
Your laptop has same serious overheating problem.

[Porter]

What does mounting debugfs have to do with passing variables to the radeon power management?

I already had him try passing dyndpm to /sys/class/drm/card-0/device/power_method with no change. Is forcing the profile to "low" the right way to diagnose this?

Thanks for pointing out that there's an overheating problem, we've figured that part out so far. The fact that it doesn't overheat in Windows rules out a hardware problem, though.

[Schmaker]

It shouldnt be really hardware related .. i had same problem with HD2600 - overheating is solved by driver by tuning GPU power management to the way laptop can handle with its limited cooling.

[darkbasic]

What does mounting debugfs have to do with passing variables to the radeon power management?

He needs it to see if it actually lowered the frequencies.

Is forcing the profile to "low" the right way to diagnose this?

Yes, because dynpm still sucks hard.

Thanks for pointing out that there's an overheating problem, we've figured that part out so far. The fact that it doesn't overheat in Windows rules out a hardware problem, though.

No, a laptop shouldn't overheat even in full load so it only states that his laptop is full of dust or badly designed.

[darkbasic]

Yes, because dynpm still sucks hard.

FUCKING EDIT TIME LIMIT SUCKS EVEN HARDER

[Porter]

With profile forced to "low", temperature is holding at 89C. That is still really, really high.

This does not occur in Windows, and the heat pipes and fans are working. Dust free.

Any other ideas?

[darkbasic]

Then it isn't the vga, you need to undervolt and underclock the cpu with cpu-freq:
cpufreq-set -c 0 -g powersave
cpufreq-set -c 1 -g powersave
...

where 0 and 1 are the cpus.

[darkbasic]

Dust free.

Unfortunately that means a badly designed laptop or too much termal paste. You will never be able to use its full power without overheating, you can just keep the temperature down with powersavings.

[piete]

Unfortunately that means a badly designed laptop or too much termal paste. You will never be able to use its full power without overheating, you can just keep the temperature down with powersavings.

That may be true if the OP's friend has played 3D games on windows and the computer still overheats. However, to see if you can use any linux on it, try installing Debian Lenny (the oldstable version) with ATI's proprietary drivers, or even better for testing, try to grab PCLinuxOS 2009 live CD from somewhere, after it starts just type "sudo aticonfig set-powerstate 1", and see if your laptop still overheats (it should not).

As good as the progress with OSS Ati drivers is, the driver (or whatever component it is today that does the power management) cannot keep my Mobility X1600 at its lowest temperature for a prolonged time (but it won't let it overheat, either). That's why it's still Debian Lenny for me.