Back again, this time with offset voltage and a 16 hour prime blend run at 4.8. I noticed that what the VID in realtemp was showing was almost exactly what I need for stability at 4.8. I ran offset with +.005. I wish there was an offset option for +-0.000.
XMP set in BIOS with BCLCK at 100 and memory voltage at 1.25v
CPU Core offset voltage of +.005
PLL set to 1.700
VRM Frequency at 350
Duty Control on Extreme
Phase Control on Extreme
CPU Current Capability at 140%
LLC on Ultra High
Pretty much everything else at default auto settings
During the run, VID would vary between 1.3511 and 1.3561. CPU-Z showed voltage anywhere from 1.344 to 1.352 to 1.360 (Screenshot shows 1.352). At idle after the run, CPU-Z showed 1.008 and 1.016 while VID was 1.0107. I haven't surfed around the web or anything like that on these settings yet, but I'm hoping for no idle BSODs given those voltages. In BIOS at full stock settings, the CPU vcore is usually 1.120.
After all this, I made one small change in BIOS just to see. I set voltage to offset and instead of +.005, I set it to auto. The machine booted at 1.54 vcore in CPU-Z at the Windows desktop! I quickly took care of it. Anyway, off to the screenshots.
BIOS settings first:
16 Hour Prime Stable
Temps 64 73 71 68
Sorry for the gigantic post. Next step is to see how far, if at all, I can lower PLL and still be stable. Then I'm going to work on a 24/7 4.5 or 4.6 OC, haven't decided which to shoot for yet. Whenever I get some watercooling, I'll probably see what it takes to make 5.0 stable.
I previously tried to get 4.8 stable with 1.355 manual voltage, but couldn't make it through Prime 95 for 12 hours. Got a BSOD 124 after 3 hours. I upped VCCIO slightly and 1792 failed in 3 minutes. I set VCCIO back to auto and raised RAM voltage to 1.275, but got a 3b BSOD after 25 minutes of 1792. Therefore, I feel like I've gotten vcore as low as I can for 4.8.
Thanks for all the good info in this post. Hopefully, something I've shown here is useful to someone.