Duron's temperature

[Stephen24]

My Duron 700 at 970 1.85V is 40C full load (prime 95) With artic silver thermal paste and a FOP 32. I am really happy with my FOP 32 right now. Also, I have good case cooling. Case temp is 27, goes lower when i turn on all the fans but it's loud I turn them on if i am doing some intense gaming.

 

[Priit]

Mikewarrior2: Is there any way to get more accurate temperature of socket A CPU than +/- 10C ? Borrowing some high-tech thermistor and trying to measure by hand ? In my case, I understand that I SHOULD ( ) be O.K., but the temperature is still bit too high ? Too bad that the other computer components can't survive -15C, putting the box behind window in winter should guarantee pretty cool CPU.

Without any software cooler, Win9x AFAIK doesn't use HLT commands when CPU is idle, so CPU's temperature shouldn't be much different betweet "full-load" and nothing in that case...

 

[Mikewarrior2]

Priit,

Sometimes it isn't +-10C. Sometimes the Mb is good at "approximating" cpu temp, like Balael's case. Since you're using an A7v 1.02(with socket-thermstor) and a 1004 or later bios, your mb tends to approximate temps a bit higher than core(tends to, i've seen it under-approximate, too, but usually at high overclocks).

Keep in mind that if balael were using an older bios, he would be reading cpu temps in the 20C range. Gee, I wonder why they implemented compensation in their BIOS'

An external thermsitor is better. It still isnt' perfect, but its the most-low-cost way for a socket a setup's temps to be measured accurately. You'll get some temperature compression, but not as serious as on a socket-thermistor setup.

And to touch-base on the socket-thermistor compressions, here's a bit more info:

Amd rates Ccore to Cback has a resistance of .5. So in the very best case, a socket-thermistor will touch directly behind the core, and is isolated from socket-air. With this setup, you'd be looking at a 2X compression when measuring temps versus Core TEMP changes. But this isn't reality. No mb touches the center back of the core, nor are they isolated. From tests, we know that the kt7(whose thermistor touches backside of cpu core edge) measures roughly 25-30% lower than than that of measurements directly behind cpu core. SO, you're looking at, 37-35% of core temp heat measured at kt7 thermistor spots. This is ignoring air interference, which on some heatsinks, is worse than on others. Knowing that at times, you get a 5-6x Compression factor, there are instances where the thermistor only registers 16-20% of cpu core temp change. So, i think htere is a minimum 2x compression of Temp changes, all the way up to a 6X compression.


Mike