Figured I'd make my own thread and stop cluttering up another. So, long story short, I got a 3700X to replace my 1800X. PBO and core sleeping doesn't work right since I have an X370 board, it'd top out at 4.3ghz even under just a single core, so I tried manual overclocking. First I tried 43x100 at 1.4v, which resulted in...yea, didn't last long, thermal shutdown right after this screeenshot was taken.
So after numerous trials using lower and lower...and even lower voltage, and switching to P0 state overclocking, I arrived at this: 43x100 1.175v, which by all accounts shouldn't even POST much less pass an AIDA64 cache stress test for over half an hour (manually stopped) and function perfectly fine in every program I throw at it.
Now, it's not Prime95 stable, it crashes in test 3, so it's not an -absolute- stable system, but both my 1800X and FX-8350 failed Prime95 as well, but who cares about that aside from the number crunchers trying to find a prime number? It's a heck of a lot better than AMD's default voltage of 1.45, that's for sure.
Kinda disappointed in the (rather meaningless) 3DMark score, but if my CPU were anywhere near unstable, it wouldn't pass. Note I switched back to using the regular overclocking vs P0 overclocking, for some reason it would actually start acting up under a long duration AIDA64 stress test or hours of encoding with Handbrake when using P0 but not when doing it regular. Only difference it makes is that the voltage is now read incorrectly in Ryzen Master and 3DMark, no big deal.
Slight update. I added 2 more 8GB RAM chips, what can I say, stimulus money. Anyway, other than that the other specs are the same. Cinebench score increased to 5028, a large increase outside the margin of variability, which does backup the claim a couple of reputable review sites made that Ryzen chips benefit from all four RAM slots in use. It had a negligible effect on the 3DMark score (not a surprise), but the CPU score increased by 350 points.
