Originally Posted by miklkit
Your 8320 sounds a lot like my old 8350. The most I could get out of it was 4.8 and I pushed it as hard as 1.6 volts on the Sabertooth. I used regular LLC and vcore would vary from a high of 1.608 to a low of 1.524 to 1.536 at 100% load. If your vcore is going up at 100% load, then you are using too much LLC in my opinion. The most that others have used keeps the vcore steady at all loads while I accept a little vdroop in the interest of lower temperatures.
Stress testing is important, and not all tests are created equal. As I see it IBT AVX comes closest to real world loads what with games pushing the cpu harder than before. The argument that you never see loads like that are not as persuasive as it used to be and you have to know your system is stable at high loads and temperatures.
Yeah, my vcore is going up under load, first to 1.548, and after some tome it gets to 1.56V. I thought it would be better to have stability under full load, never thought idle would crash it. I was stuck on constant 1.60V for some time on my DS3P and forgot a few things. Seems like I'll need to try 1.548V at idle and 1.56V under full load to maintain stability.
Biggest problem now is that stress tests pass and full load isn't the issue, it's the idle state that messes with the system. But with LLC messing with idle voltages too, it's kind of hard to get the needed LLC level and voltage setting in BIOS to get that 1.548-1.560 under load.
This all started when I was setting the pump to work at 50% PWM when under 40C socket temps. At one point I saw it was 0RPM and the core temps were 70c, I powered the system off imidiatly, and I had some trouble booting then, did the CMOS reset, flashed the BIOS again, but I'm getting those weird things happening on my system now that didn't happen. Either the chip really degraded or it's the combo of a bad chip + wrong power settings in the BIOS. Anyway I'm getting pretty annoyed with this.