[ltp] Everlasting Fan

[Alex Deucher]

On 6/14/05, Ovidiu Achim <amorox@gmail.com> wrote:
> Hi,
>=20
> I tried to look for it on dri-devel and xorg without success. Google
> gives no hint also. It would be nice if you remember where you have
> seen it.
>=20

er... sorry, it looks like someone mentioned it to me in a private
email.  I'll ask him if he plans to release the code.  If you are
included to play with it yourself (be careful), you'll want to
manipulate the XCLK and MCLK.  take a look at radeon_driver.c in the
xorg ddx to see how clocks are normally set.

Alex

> Thanks,
> Ovi
>=20
>=20
> On 6/14/05, Alex Deucher <alexdeucher@gmail.com> wrote:
> > On 6/14/05, Thomas Hartwig <t.hartwig.news@itth.com> wrote:
> > > Hello,
> > >
> > > unfortunately ATI still makes no advantages on linux, so I'm wonderin=
g
> > > if anybody can give some deeper help on debugging the xorg provided
> > > radeon driver.
> > >
> > > My experiences are that the temperature of the GPU is the most bigges=
t
> > > problem related to the fan for my T40p. Typical values are:
> > > CPU: 44=B0C GPU: 52=B0C HDD: 32=B0C PCI: 42=B0C. I have decided to tu=
rn off the
> > > fan below 50=B0C, but independent of the output of X or console the
> > > temperature is always at about 50=B0C or higher for the GPU.
> > >
> > > The radeon driver has the nice option "DynamicClocks" which might alt=
er
> > > the power of the display. I'm wondering now if you can fine tune this=
 in
> > > a better way and see what is going on. I have already taken a small l=
ook
> > > into the source but I'm afraid without much time I'm not able to
> > > disassemble this ;-)
> >
> > The dynamic shifting of the clocks is handled by the hardware, itself.
> >  I'm not sure exactly what its algorithm is.  Pretty much all the code
> > does is enable the right combination of enable dynamic clocking bits
> > in the various clocks.  BTW, the code has some bugs in it that I need
> > to fix at some point.  Also just to clarify, this is not powerplay.
> > Powerplay is more involved and ati hasn't released the information
> > needed to do it.
> >
> > > What would be helpful is: at what clock the GPU is always running and
> > > what the strategy for the regulator is.
> > > Is there anybody out there who might be interested in looking at or h=
as
> > > already done?
> >
> > A while back someone posted a utility (maybe on xorg, or dri-devel)
> > that allowed you to statically adjust the GPU clock rate.
> >
> > >
> > > Perhaps I will contact the author of the code as well.
> > >
> >
> > Ati wrote the code, I just cleaned it up a bit and committed it to cvs.
> >
> > Alex
> >
> > > Best regards
> > > Thomas
> > >