system temperatures still high after stress testing with Prime95

[Dubslow]

Um... no. Absolutely not.

Power is a specific measurement which is completely different from temperature, and even then the Thermal Design Power isn't actually the maximum power that a processor can consume. Furthermore, just because one processor has a lower TDP then another, does NOT mean that it will be cooler. In particular, Ivy Bridge quad cores, despite having a 77W TDP, tend to run hotter than Sandy Bridge quad cores, which have a TDP of 95W.

A Pentium D's TDP is 95W, not 95C. The former is a measure of energy per time, while the latter is (sort of) a measure of energy per entropy, entropy being (sort of) a measure of "disorder".

It's Thermal Design Power, not Thermal Design Temperature.

[Batalov]

Originally Posted by Dubslow

Um... no. Absolutely not.

It's Thermal Design Power, not Thermal Design Temperature.

The first and last line would have been enough, methinks.
"energy per entropy", and the rest of the sentence! Loved it!

[LaurV]

Almost all of these new toys have a junction temperature of 100C, or even higher as 105C and 110C for mobiles (the reason is the mobiles lack big coolers, just now we are designing an industrial PC with i7 620M, whose J-temp is 105C, with passive cooling, no fans, just few aluminum blocks and heat pipes - the device is supposed to work well at CPU maximized in a 60C environment - this means room temperature. With preliminary testers we are well under 90C when P95 runs 4 workers, 2 phys plus 2 logical).

You can't fry these CPUs by running them at 90C, assuming you can constantly cool them and you don't have heat spikes which go much higher (this can not be detected by software like RealtTemp or so). New mobos have hardware which cut the power when the CPU enters thermal lock or throttle. The CPU is sending periodical signals to the mobo as long as everything is fine, independent of your software or OS. When the CPU is too hot and it enters throttling - all of the new CPU do this - the frequency of the signals to the mobo decrease. Clever mobos sense this and cut the power.

Asus will give you all money back if you can fry a (for example) Maximus Extreme by overclocking.

[kracker]

Are you saying it's fine to run them at 90C? I realize mobile cpu's can tolerate more heat, but I like to keep mine at about 50-60C (lower if possible)

[LaurV]

Originally Posted by kracker

Are you saying it's fine to run them at 90C? I realize mobile cpu's can tolerate more heat, but I like to keep mine at about 50-60C (lower if possible)

Not really. To keep the temperature as low as you can would be your goal. This will extend the lifetime of all your hardware, not only the CPU. See my older posts where I was talking about "bent plastic parts" and "fan blades deformed from heat" in 2-3 years on Sony laptops. Higher temperature shortens the life time of the hardware, not talking about additional energy consumption. In real life, with real tasks and real coolers, the temperature is never "constant". There are always up/down spikes which you can only measure with professional hardware, like in an electronic factories (I work in an one). To get a real feeling of what's happening, set the throttle=50 into prime.txt, and run for 10 minutes and look at the temperatures recorded by RealTemp. That is what some professional measuring device shows when there is no throttle and P95 runs at full capacity and RealTemp shows a constant temperature. Measuring equipment will always show oscillations of few degrees up and down even if when you think the temperature is constant. My point was: if you run your CPU at 90, for short periods of time you might have (undetected by the measuring software, which does averaging) 80, but you also might have 100. If you can keep the temperatures low, then better do so.

[kracker]

Originally Posted by LaurV

Not really. To keep the temperature as low as you can would be your goal. This will extend the lifetime of all your hardware, not only the CPU. See my older posts where I was talking about "bent plastic parts" and "fan blades deformed from heat" in 2-3 years on Sony laptops. Higher temperature shortens the life time of the hardware, not talking about additional energy consumption. In real life, with real tasks and real coolers, the temperature is never "constant". There are always up/down spikes which you can only measure with professional hardware, like in an electronic factories (I work in an one). To get a real feeling of what's happening, set the throttle=50 into prime.txt, and run for 10 minutes and look at the temperatures recorded by RealTemp. That is what some professional measuring device shows when there is no throttle and P95 runs at full capacity and RealTemp shows a constant temperature. Measuring equipment will always show oscillations of few degrees up and down even if when you think the temperature is constant. My point was: if you run your CPU at 90, for short periods of time you might have (undetected by the measuring software, which does averaging) 80, but you also might have 100. If you can keep the temperatures low, then better do so.

Ah, I see what you mean. Thanks

[davieddy]

Originally Posted by Batalov

"energy per entropy", and the rest of the sentence! Loved it!

At the risk of repeating myself,
Shannon asked von Neumann what he should call his new information theoretic expression.
"Call it entropy" came the reply.
"After all nobody knows what it means anyway"

David