Originally Posted by GigaByte
CPU temp is not measured by a mobo "sensor", the signal is taken from the CPU diode itself and converted by the super IO chip and northbridge. Intel has stated the accuracy of the thermal diode (Tcase) is +/-1c and the thermal diode is indeed between the cores at the middle of the die. You are correct about the 5c delta, but that is on small FFTs/blend on Orthos or Prime95. Intel's Thermal Analysis Tool (TAT) is made strictly for mobiles, as we all can see our desktop Core 2's are seen as Pentium M's. TAT has given the weirdest of all temps for many users, some get very low and some get very high and full load on TAT is equal to 114% load of Orthos or Prime95 on small FFTs. The delta you should see with TAT is typcially 8c and in many cases higher.
I think you made a mistake on the last part, CPU temp is the Tcase, you seemed to have said CPU temp and Tcase are different. There is only 2 temps, Tcase and Tjunction, Tjunction means core temps. I am not trying to bash you but if you think Tcase is wrong or unimportant you are going head to head against the company that made these processors, the info for Tcase is right on their site.
This will be my last response because I tire of the effort. I am going to stick to XS, where people actually read and understand the readily available intel docs.
Tcase is only accurately measured ~380 microns into IHS via a thermocouple, period. This is specified in nearly all the intel thermal docs WITH PICTURES.
"CPU" diode temp is not measuring Tcase at that point, ever, and on load it matters. If your cpu has a diode sensor, then temp is being measured between the cores. To go from between the cores (diode), you pass through more core, TIM1 and IHS, and intel research suggests roughly 50% of resistance is in TIM1 and IHS on load, therefore you are missing 50% of gradient, therefore you are not measuring Tcase, you are somewhat poorly approximating/over estimating it.
Intel does refer to Tdiode as "approximation" of Tcase, but in no way is it Tcase, and it will read higher than Tcase. If it read the same as Tcase, intel would not go to the trouble of boring 380 microns into IHS with a thermocouple as they explain in most of their thermal papers to tell you that is the only way to accurately measure Tcase.
Intel supplies the thermal conductivity with the formulas, you can even work it out for yourself if you dont believe intel.
Why not try answering the questions below.
Cpu temp is measured where?
If cpu temp = Tcase why does intel not just use diode in place between cores, why do they cut a 380 micron slit in cpu to measure Tcase?
If cpu temp = tcase, and cpu temp is tdiode, then tdiode=tcase, so are you saying there is no gradient on high load from between cores, across tim1 and across IHS?