On 30 Aug 2004 22:04:52 -0500, wrote:

IMHO they start throttling after 100°C from the onDie diode ...

I think you are mistaken. According to the Intel product documentation the
maximum operating temperature for the Xeon is listed as 71C.

since the introduction of IHS on their CPUs, IMHO Intel gives specs
for max. long term full load !!! reliable operation temperature
measured by the temperature of IHS. That means that onDie diode
one is at least 10°c higher @ idle & 15° higher full load. That means
that if CPU fan dies , you have 70° on the bottom of the HSF (IHS),
but inside the core /cpu diode/ you have there @ idle 100°C (which BTW
is also declared handling/storage max value) & IMHO thermal throttling
should not kick_in much before that to slow down the sistem.
Thermal throttling will keep Cpu die below 120°C (remember
that famous video with running it w/o a HSF & not crashing OS!) if HS
looses contact (or badly mounted) with IHS ... Second thermal
protection incorporated onDie, will shut down CPU at IMHO if reaching
125°C /quite general rule; in industrial inviroment that temps are
generally higher/.
On some MoBo´s, you can set also inside Bios thermal
throttling at dedicated temp (also warning & shut down ones too).
That for precaution or a thermal runaway I´ve done for my Tuallie when
I was for testing running it w/o a fan OC-ed in closed case with a bit
bigger HS than original. (max I was running it max throttled by mine
MoBo @ almost 90°C from onDie measurement w/o crashing my OS - but was
slow as a dog!) & IHS temp barely rised over 60°C !
Check my site under comp/benches to see some mine temps with
different fan speeds & imagine P4 /or Xeon/ temps than while it
consumes more than double power than my Tuallie ... (my temps are
correct ones - sensors properly calibrated!)

about GPU:

IMHO its HS temerature should not exceed 80°C (means that
inside GPU would be around 100°C even if Fan cooled HS ...
mine prehistoric video card /passivelly cooled with added bigger HS/
craps out after 70°C HS temp (inside GPU there could be same around
100°C)

That my CPU fan experiment there I mentioned before with
lowering fan RPMs & consequencies can be applied also on graphics card
cooler when lowering voltage of a GPU HSFcooler avoiding loud
"turbine" sound coming out from the PC case ... :-)

the real problem is to exactly know, how the temps on CPUs & GPUs are
measured by the certain MoBo/graphic card - where & what kind of
sensor is used while the monitoring program is showing temps
even properly calibrated & set ... (or some latest drivers have
interface for GPU tweaking & monitoring temps of GPU, but has to be
verified what kind of sensor is used anyway on a graphic card & where
is situated ...)

About setting the alarm temperature for GPU :
I would set it a degree or two less than running for awhile Doom demo
in loop on the hottest summer day ... before it freezes the PC! ...

sorry for the long post; but I think it could be informative & useful!
--
Regards, SPAJKY ®
& visit my site @ http://www.spajky.vze.com
"Tualatin OC-ed / BX-Slot1 / inaudible setup!"
E-mail AntiSpam: remove ##