View Single Post
  #4  
Old December 19th 03, 09:56 AM
Kevin D. Kissell
external usenet poster
 
Posts: n/a
Default

"Bitsbucket" wrote in message s.com...
Header says it all, I overclocked my 2500+ Barton core to 3200+ speed by
upping the FSB to 200MHz. This results in about a 5 to 7 degree C increase
in diode temp. Should it go THAT much higher? I am using an Antec fan I
picked up at Circuit City, has a copper plate for the bottom and VERY fine
fins, fan around 2800 rpm.
BUT it dosen't matter if I use that fan or a MONSTER solid copper job with a
fan from hell, (read LOUD) runs about 7K rpm. I just can't figure out how to
get the temps down, Vcore is stock, cover is off the side, 1 incoming fan
and 3 exhaust fans. When I put my hand near the CPU I can feel the heat!
There has got to be something better, I know I'm negating the intake and
exhaust fans with the cover off, but I don't think it will be any better
with the cover on (or maybe it would?????) ambient temp is around 70 deg F
in the room, celing fan running so there is plenty of air movement in the
room...........any ideas? I just set the clock back to 2500+ and now I'm
running 45 to 47 deg C on the diode........at 3200+ it runs around 49 to 52
C (all temps are at idle, unless I run something then it jumps to the higher
numbers)
What is too hot? 60C is what I'm thinking that is why I don't like those 52
and even some 56 C when something intensive is running....


Processors running in the 2GHz range get hot, period. Depending on the model,
AMD rates their CPIs to 70C or 80C. Your diode won't necessarily be providing
exact data, so you want to leave a decent margin. Running at an indicated 60C
is probably not dangerous, but plainly, the cooler the better.

The power dissipated by a chip is the product of the frequency and
the voltage squared. You can't take a chip from 1.83GHz to 2.2GHz
at the same supply voltage without expecting something like a 20%
increase in peak power consumption, and a consequent increase in
heat dissipation. Because of the exponential effect of voltage changes,
reducing the core voltage by a small amount can bring about a substantial
reduction in power dissipation at the same frequency, or allow you
to overclock and dissipate the same amount of power as at normal
frequency/voltage. The catch is that reducing the voltage increases
the probability of the logic circuits latching erroneous values, causing
Very Bad Things to happen to your computer - if you're lucky,
it will crash before anything corrupt gets written to your hard drive.
In my own limited experience with overclocking, I sometimes had
to *increase* the supply voltage (and thus the heat dissipation) to get
stable operation at higher frequencies.

If you can't reduce the amount of heat generated by your CPU,
you need to look at how to get the heat out of it more efficiently.
The choice of heat sink/fan can make a huge difference, as can
the care with which you mount it and the kind/amount of thermal
paste used. Even the best thermal paste conducts heat less well
than even the worst heatsink material, *if* there is contact between
the chip case and the heatsink. Paste is necessary because neither
the chip casing nor the heatsink are perfectly flat, and you need
to fill any cavity with something that will transmit heat better than
air. There are those who sell silver and copper-based pastes which
have higher heat transfer rates than the basic silicone-based pastes,
but the metal-based pastes can form thicker layers than silicone,
and a very thin layer of silicone will conduct more heat than a thick
layer of silver stuff.