Set GDM (GearDownMode) to off
FCLK either 2000 or 2200 (you have to test this, preferably in game & with Prime95 Large FFT (AVX on)
Hi guys, agreed that GearOFF & FCLK 2200 all day is the way, best route to max. performance.
About testing stability in games, nope never was the guidance way at all, Prime large FFT always fits better to Intel cpu behavoir scenario, for Ryzen do you need Y-cruncher FFT/N64/VT3 combo best shoot & hardest to pass I've throwed up so far to Ryzen cpu's era, it will throw errors as soon as your memories heat up high enough to keep it up with IMC controller in stable AVX2 testing scenario hopefully it is just a voltage issue that it takes care of or an AVX2 test will detect it. coming soon 👍
I know it sounds counterproductive, cause OC memories hard needs a lot of time consumer, doesn't matter 1 #error or 10hours TM5 run, wrong timmings, low VDD or Q voltages, maybe poor cpu IMC or simply too much PCB sticks high temps heating them up quickly idk, so confident y can find out which one y issues are 🤔
But although I trust the Y-Cruncher test suite which has passed me well a few times now, AVX2 dataset is hard, so will N64 find out memory controller errors instant by cores 💪
Best to test with Karhu, TM5 often cheats anyway, you just don't see the test speed... (or use 0.12.3)
@Gr3yh0und TM5's job is just to check the memory timings, nothing around it, but it will also fail if the CPU can't maintain stability.
TM5 is not able to detect memory control issues, for such work you need Y-cruncher, it can find out in no time memory controller errors by cores weaknesess instant.
Karhu test is good too, coz uses different approach kinda testing much more the cache side part of things and stuff.
Karhu with cache will fail if the IMC fails to hold this voltage, or also will fail if one of the cores fails indeed.
I strongly recomend to try Karhu or Ycruncher for such stability testing work, keep in mind Karhu with Cpu cache=Enabled always 🙏
Karhu test and Ycruncher it's fit better for everyone who tested
only TM5 stable so far, and they are already validated Profile setup as stable, but guess what? maybe you are not stable as y think y are?
@draytek11 please use Karhu test, better for his memory cache side of things testing approach, and than Y-cruncher AVX2 suite later if you passed Karhu 12h+ well minimum 40000+ coverage %
Hope if all fine you are good to say "hooray that's stable, no errors" endlessy.
Nothing but true, I saw same yours timings somewhere else before 😜 doing just fine, well done stable on my Neo's A-die setup indeed 😋

But actually founded better tightest timings 👍
And a bit 100.5 bclk included won't hurts 🔝