@mrthanhnguyen
Wait is this actually stable? If it's not thoroughly stable, then all bets are off because AIDA starts snorting crack when not stable and the sky's the limit.
- If stable, then that's the first time I've ever seen high speed pure 1T stable without 60 Clkdrvstr.
- If you're just chasing AIDA scores I don't feel like tRAS needs to be that low.
- For a daily-able system no point going below 4-4-16-4-8-10 for tRRDS to tWR. Looking at that tWTRS @ 3 there.
- Never seen odd tRTP before, not saying it can't be done, but tRTP 6 should always work with no problems on high end Bdie. Also some sort of weird bullshit relationship between tRTP and something else, can't remember, best to keep it 6.
I don't like the idea of the "auto OC switcher" because it just sounds like a dumb algorithm that won't respond quickly enough in some benchmarks that the board designers didn't take into account. The CPU-Z test is extremely short, might very well not be triggering, use R23 single-thread to test instead.
I have no frame of reference for R15, but 675 in CPU-Z should fall roughly around 4850MHz effective clock. If that "max" 4960MHz reading on Core 0 was during CPU-Z ST bench, then your settings clearly aren't working very well with the dynamic OC switcher. Go back to your PBO settings alone. Use Cine R23, easier to compare and get an idea of where performance is.
VDDGs are looking okay, VSOC could be a little high, but VDDP is way too high. Drop that down to 0.9V or 0.95V unless it just won't work without touching VDDP.
Don't get too ahead of yourself. Make 14-14-14 work first before you start dropping tRP and tRFC like a stone.
View attachment 216654 View attachment 216655