Originally Posted by Flaxen Hegemony
Delidded on AIO. The delid showed temp drops of around 10C (34 ->24 at idle), which seems to be above average for this chip. I attribute this to the Intel paste job being particularly bad, rather than any delid expertise on my part.
No throttling is happening after I reduced voltages. At "Auto" voltages and 44x multiplier, HWINFO showed VRM throttling after the VRM temps hit 109C. In Prime95 with AVX enabled / no offset, CPU was throttling shortly after VRMs. This all occurred within 2-3 minutes.
After setting volts to 1.15, there was no throttling. VRM temps were uncomfortable at 85C, but CPU cores were far from problems at 75-80C.
In case it interested anyone, I ran some statistical analysis on the temperature and voltage data collected, and found the following:
Correlation between idle voltage and load voltage on auto was .013, no relationship. A core needing less voltage at idle won't need less at load.
Correlation between load temps and load voltage on auto was -.011, no relationship. Across cores, voltage isn't alone in creating heat. Other factors at least equally important.
BUT, there was a correlation between idle voltage and load temps, a significant one at r= .450. Higher idle voltages meant higher load temps.
It seems odd for this pattern to occur, unless perhaps voltage needed at idle and load temps are both related to physical properties of the core, whereas voltages at load are due more to the BIOS response to a multiplier, i.e. overvolting. If this hunch is correct, then one could likely create even more power and temperature headroom by tweaking voltages at the core level. This would take a long time, and be tedious, but the result would be likely worthwhile.