[ltp] An "inverse" fan problem

[Michael Keyl]

On Thursday 26 May 2005 14:06, Jan Girlich wrote:
> Michael Keyl schrieb:
> > On Wednesday 25 May 2005 02:04, Jan Girlich wrote:
> >>What about posting this small test-program? If you tell me how to use it
> >>and how you did the temperature-check etc., I will test it on my R51
> >>1.5GHz.
> >
> > The following C-program can reproduce the effect, although it takes
> > longer to reach the passive temperature (but the code is not very
> > efficient of course). I have monitored the temperature as before.
>
> I tried it with this program for about 1 hour. The maximum temperature I
> got was 65=C2=B0C after 670 runs and no change in the cpu clock.
>
> Maybe this program doesn't use the processor as hqrd as needed to
> reproduce your problems. :/

Thanks a lot for your efforts! This supports somehow my own observations. A=
s I=20
have mentioned already in my last mail, I have experimented a little bit wi=
th=20
lower frequencies, and the highest clock below the max of 1.7Ghz, which I c=
an=20
set is 1.4 Ghz. At this clock I am not able to heat my cpu above 76C. Not=20
even with intensive use of the ATLAS library, which brings _at 1.7Ghz_ the=
=20
processor to 92C within 2 min. BTW on my machine the cpu temperature seems =
to=20
be almost irrelevant for the fan speed. Its mostly triggered by gpu=20
temperature.

Together with your result, this indicates that all R51 models share the sam=
e=20
fan model (and the same control program) which is sufficient for 1.5Ghz=20
processors but not for faster ones. Would be interesting to know, what=20
happens with Dothan processors, which are available AFAIK up to 2Ghz.

Well, all in all, the problem seems to arise only with extreme application=
=20
which uses the fpu a lot (even the ubench benchmark site runs through witho=
ut=20
trigering passive cooling), or which uses high cpu and gpu power together=20
(i.e. opengl games). This is something I can live with (although its annoyi=
ng=20
because I do some numerics from time to time), but, on the other hand: it=20
can't be very difficult for IBM, to equip each notebook with a sufficiently=
 =20
powerful fan ...

Michael
=20