LOL Virgin Mary 2. "Because if she were any tighter, she'd be a guy!"
So I've gone back to the start, working on getting 3200 to pass AIDA Stress on Auto everything except GearDown Disabled and 15-15-15-35 1T. RAM is set at 1.35V (1.36V actual) and CPU-NB is set to 0.960V (I get the feeling the board reading is wrong and I need to look at what the Ryzen readings show which is 0.956V and 0.950V under load [same as the MOSFET reading]). Only other change was I looked at the LLC stuff and I disabled the "DRAM Phase Control" option which says it would adjust the phases in use based on the DRAM Load, which I figured may be beneficial to my cause (I've been running with the Power Control set to "Current Balance", and left that set as well).
However, I still don't like what I'm seeing in HWiNFO's sensor readings... DRAM Voltage generally is 1.360V, but occasionally will dip to 1.344V for a moment, and I get the feeling that can't help matters.
Just ran the stress test and it still fails *sigh* Guess I'll try increasing CPU-NB a bit more. There aren't any WHEA errors though, so I assume it's not memory/timings related?
Anyways,that's not what I'm here for really... I've still be looking into Windows 10 Timings and after running for about a week with HPET on in the BIOS but off in Windows with the Perf Timer basically always reading 3.320MHz... something odd happened when I turned it back on just now and rebooting. Now it reads 14.318MHz! o_0 I find that odd since before turning HPET off in Windows and BIOS, it was reading 3.580MHz. If this holds up and is the same every reboot then I'd have to conclude that for whatever reason, the Ryzen Master batch file that turns it on, hasn't really applied it (despite confirming with BCDEdit that it is infact turned on). *shrug*
I also found "WinTimerTester v1.1" which will tell me if it does in fact change (since I'm not really trusting CPU-Z's Report txt), and allow me to test to see if things are still in sync after using K17TK (referring to timing drift). Currently everything is good, but I haven't changed clocks, so I'd expect it would be. More importantly are benchmark results are the same between whether HPET is Off or On in Windows (at boot-clocks).
EDIT: Welp, color me confused. After setting P0 FID to 38x, there's no timing drift, all the timings show normal (the same), yet AIDA's scores go out of whack (which doesn't happen with using Ryzen Master to change the multi). Pi though? It actually doesn't change. It performs with the same speed with FID @ 39x as it does after booting with the CPU Boosting to 3.9GHz. Makes me want to conclude that Windows is fine, it's the specific software that are somehow at fault. I CAN say that if I change the Timer Resolution from 15.6ms to 1ms, then my Pi score changes (becomes slower actually, by about +0.600sec), but AIDA doesn't. So I officially give up trying to make sense of that
Bookmarks