According to Intel, this is not a problem with the CPUs. It's an issue with immature BIOS/temp software reading improperly. We've taken returns on chips with this problem, yet when I pop them into our test machine, they work fine.
Hopefully intel was speaking of cpu temps, which are indeed dependent on bios calibration. It would be very disappointing if intel stated core temp problems on the E8400 were either a bios issue or a software issue, as that is false.
The delta to Tjmax/Prochot# (processor hot #) is accurately reported by coretemp, realtemp, speedfan, etc if they are configured to read "delta to tjmax". They are all accurately reading this value directly from the digital output of the DTS on the cores, the bios is irrelevant as is the particular software used to read the value.
It is true the reported
absolute temps based on DTS reading will be reported 10C lower by realtemp (guesses tjmax 95 based on fluke measurements) than coretemp/everest/HWM (guesses tjmax 105 because 45nm mobile cpus use tjmax 105, even though intel said that was not correct), ie, absolute temp = tjmax (guessed) - DTS. However the actual DTS reading on cores of mormonsniper's E8400 are showing
10-14C difference between two cores. That is 100% an intel issue.
Either one core was individually calibrated to throttle 10-14C below the other, or more likely one or both of the sensors is no where near the claimed +/-1C accuracy, which is apparently a problem that was magnified when moving to 45nm. Though rarely some of these sensor discrepancies have been reported to improve by reseating or altering heatsink pressure, the vast majority report the exact same temp problems over multiple reseats.
If intel wants to argue DTS are still accurate enough for throttling thus protecting the cpu and that is the only issue, that is their prerogative, but misinforming is not.
As an aside, I dont think the E8400's are limited by temps, as the voltages are likely to be a limiting factor before problem temps are reached.
One can still get approximate temps if use your cpu temp as a guide. And that temp, you will need to update your bios. Then set your E8400 to .9v and 6x200, and calibrate cpu temp at idle in speedfan (latest 4.34 beta not 4.33) to +1 above ambient. Then put back to stock/OC settings and run stress test to get E8400 core temps in higher range where they are usually more accurate, and core temp should now be 1-5C above cpu temp.
http://arxiv.org/ftp/arxiv/papers/0709/0709.1861.pdf see figure 5. That way you might figure out which of two core temps is closer to absolute.
Then we will just have to learn to live with some borked core sensors or rely on cpu temps until next upgrade ...at least by time nehalem gets here, hopefully intel will have temp sensor problem fixed, perhaps use the DTS version of something like this...
http://www.national.com/appinfo/tempsensors/trutherm.html
