OK, make that 70C for the new 5500 series processors then. Keep in mind though than even Intel gets very nervous at 75C and from that point the processor WILL go into a TCC partial shut-down. 60C ~ 65C is still a better target (of course) and should be obtainable if there's a good heat-sink and a user controllable fan.
I just went back and looked, and at spec, they list Tcase = 67 or 67.9C. So 70C is close enough, but a little cooler is better.
Examples:
i7-975
i7-920 (C0 stepping)
W3580
W5580 (DP example)
In the case of the units used in the '09 MP's, it's Tjunction, and not listed. But the above data should be an idea of what to expect in an ideal situation. Unfortunately, it's not that easy to do, as I'd guess those temps where determined on a bench, and in open conditions (not stuffed in a computer case). So the variance would make sense, assuming my recollection of 100 - 108C is accurate, as real world conditions rarely coincide with the benchies.
🙄 😉
As per the 100 - 108C limit, I just can't find the source.
🙁
Hi nano, I figure you mean a temp of 80c in Windows, right?
Yes.
🙂
To be specific, it was Win7 RC 7100-64 bit.
Anyway, here's my 2 cents on this... When running 3DMark Vantage in Windows my CPU hits a max temp of 76c for a second or two but it never goes over with my current fan setting. I would prefer it stay under 75c but it's not possible if I want to keep my MP as silent as I like it in Vista... I find in OSX it's a different story, I never push my MP over 65c (ever) since OSX seems to measure the heatsink temp and not the core temp. This is why I don't really believe what I see using istat in OSX. I would hate to be close to "TJ Max" and not know about it.
I understand.
🙂
Even temps taken off the CPU diodes aren't accurate (10 - 15C variance on newer processors), and need to be measured with a thermocouple. You then use the measurement to make a calibration adjustment (offset) to get accurate temps off the reading. The setting, if possible, is done in the temp software.
Real Temp 3.0 (windows) is one such example that allows you to do this.
The thermocouple is mounted dead center, and it would give you Tjunction, as it's sans IHS in the '09 MP's. As this isn't to be done without a heatsink (IHS or not), you need the thermocouple centered in a measured and machined copper plate, and TIM on both sides. Not an easy thing to come by. It usually has to be made for the specific package (processor).
So temp readings aren't accurate anyway, but a change is. That's due to the differences between real and displayed are negated when the difference is determined on the same system. You may not have accurate min & max, but delta T is still be useful, as it's at least accurate.
