I put together a P4 3.2 system about a year ago, and have had
intermittent problems with cooling, if I do any serious
computing...(i.e. Rendering a Video)...I'm not a gamer so I seldom
have that kind of a load on the system, but when I start up something
like Nero Recode to re code some Video files I have, the PWM Temp gets
to be quite high (70-80 C) and I abort it.
Research seems to indicate that my ABIT AI7 MB, although offering
"Prescott Support", does not really have very robust support for the
Power hungry Prescott. Has anyone had any experience wait a Socket 478
Prescott capable Motherboard, they would care to recommend????
A good article on the problem I am Seeing is located at:
http://www.overclockers.com/tips1179/
(I am not affiliated wit that WEB Site, just providing a pointer to so
info on the subject of my post)
Thanks for your time…
Mike
In this image here, I think I am seeing a four phase Vcore
conversion circuit on the AI7. A cheap design would use three
phases.
http://img.clubic.com/photo/00041533.jpg
The 3.20E listed here has a TDP of 103W. If the VID was 1.3 volts,
then the current coming from the four phase power converter would
be 103W/1.3V = 79 amps. Split over four phases, that is only 20 amps
per phase. Your motherboard should be able to handle that. (I'm
assuming here, that you haven't been turning Vcore up manually,
as that would increase the total power. Processor power is
F*C*V**2, so power goes up as the square of the Vcore used.)
http://processorfinder.intel.com/scripts/details.asp?sSpec=SL7PN
For proper cooling, the motherboard designer assumes you will be
using air cooling, and not water cooling. If you were to water
cool the processor, there is no "side-spill air" from the CPU
fan, keeping the Vcore conversion MOSFETs cool. A person who
is water cooling, should add a fan in the processor area, if
there is a problem with hot MOSFETs.
You can get aftermarket CPU coolers, and some of those blow
down onto the motherboard. An XP-90 or an XP-120 blow downwards.
A Zalman 7000 or 7700 series might cool the FETs a bit less
- I don't consider my 7000 to be good as a "sidespill" cooler.
A Zalman 9500 blows sideways, and I haven't read any reports
of what kind of air volume it moves, or how the air is
distributed.
So maybe an XP-120 would be a thing to investigate as a
cooling solution.
Do you know where the "PWM Temp" is measured ? Have you placed
your finger on any of the Vcore conversion circuits and burned
your finger ? A surface temperature of 55C can be withstood by
a human finger for about 2 seconds. If your Vcore conversion
stuff was at 70C, you won't be keeping a finger on it for too
long.
Feeling the components with a finger, is an easy cross-check
that the measurement is giving reasonable information. Feeling the
fins on a CPU heatsink doesn't tell you too much, as the temperature
drop from the silicon die to the fins is too large, for a
finger test to give you meaningful results. But the fins should
be warm to the touch - cold fins on something, could mean the
heatsink is not making good contact with the thing it is
supposed to be cooling.
For internal processor temperature, the Intel processors have
a throttling feature. If the processor itself gets too hot,
the processor responds by reducing the computing rate. This
happens at 65C-70C or so. This article discusses and illustrates
throttling in action. If you own a manual fan adjustment
device, like a FanMate2, you can slow the fan down, and
the point at which the processor starts to throttle, means
the silicon die is at 65C to 70C. That gives a way to cross
check the motherboard temperature readout for the CPU.
It doesn't help with the "VRM temperature", but the finger
test may tell you whether the motherboard measurement is
just plain wrong or not.
http://www.xbitlabs.com/articles/cpu/display/oc-guide_7.html
So apply some "sanity tests" first, before rushing out to
replace the motherboard.
HTH,
Paul