[ltp] Thermal problem on R50p

Michael Keyl linux-thinkpad@linux-thinkpad.org
Tue, 31 May 2005 12:08:15 +0200


On Tuesday 31 May 2005 11:41, Andre Timmermann wrote:
> Hallo List.
>
> My R50p works very well with debian sid, but there ist one anoying
> thing: sometimes the CPU is getting so hot, that the system is going
> down automatically.
>

One Week ago I have startet a thread about a similar problem.

http://mailman.linux-thinkpad.org/pipermail/linux-thinkpad/2005-May/026609.=
html

You will find some information there, but not a complete solution. I guess=
=20
your machine has fairly fast cpu like a pentium-M 1.7Ghz or faster. The=20
problem is most probable an underpowered cooling system.=20

> After 5 Minutes I can hear the fan and it seemes that the temperature is
> quite stable at 91=B0C (the trip point for shutting down is at 94=B0C).
>
> Then suddenly the temperature is going down:
> 10:57:01  temperature:             84 C
> 10:57:03  temperature:             77 C
>

This means the passive temperature is reached and the cpu clock drops.=20

> And after a few seconds up again:
> 10:57:28  temperature:             90 C
>

This is a little bit strange. You should check not only temperature but  al=
so=20
the cpu clock, to see whether its going up again or if the processor is=20
becoming too hot also at a lower clock. If cooling is insufficient also at=
=20
low clock rates you might check whether your machine is draggled inside=20
(dust) such that the air from the fan does not reach the processor (but I=20
have no experiences at all with this point, its just a guess).=20

> Unfortunately there is no way to see the fan and it's speed via the
> proc-interface. So I am not able to control this behaviour.
>

Have tried you the ibm_acpi module? You get info about the fan speed and mo=
re=20
temperatures among others. The driver is included in 2.6.10 kernel (I think=
)=20
but not a very recent one. You should check the web page at

http://ibm-acpi.sourceforge.net/

and load the driver with the experimental=3D1 option.=20

As a possible workaround for your problem you should reduce the cpu manuall=
y=20
(via the files in /sys/devices/system/cpu/cpu0/cpufreq) to a value which=20
prevents the cpu from overheating during processor intensive tasks. (on my=
=20
machine this is 1.4Ghz instead of 1.7Ghz).  I hope this helps a bit ...

Michael