After finding theses settings and giving them to friends and it working I figured I’d share. This isn’t an efficiency guide as I don’t pay for power my work does.
I have 2 Vega 56 and 1 Vega 64 with Samsung memory running at this stable if kept below 70c tmem. This might be limited scope as my 3 cars are 56 sapphire reference, 56 sapphire, pulse and 64 sapphire nitro+. But I’ve had 3 friends use these setting all with success on their array of Samsung memory cards.
In the pictures GPU 0 is reference 56, GPU 1 is 64 nitro+ flashed with the bios of my reference 56 card. The bottom is a pulse 56 with it’s default bios. The reference card is my best one and can actually run these setting with rcdrd 12 and men at 1100 for 59 something but I prefer the continuity. Switching rc to 37 from 36 and rcdrd from 14 to 13 has similar hasrates, but the tmem became harder to keep below 65c which is my personal goal but like I said anything below 70c seems fine for this. Ok settings time
amdmemtweak --CL 20 --RC 36 --RP 14 --WR 14 --CWL 8 --FAW 12 --RAS 22 --REF 65000 --RFC 249 --RTP 5 --RRDL 6 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 14 --RCDWR 12
All them have original thermal paste and I’ve run the different settings for 7 days at a time on hiveon.
I’m not shooting for efficiency here because power is free for me. But each card required different core voltage and two of them required more mem voltage, which I thought didn’t do anything but it stopped the crashes so my assumption must have been wrong. Best I was able to find the core and men voltages for each fast using the --high_sample_mode=8 argument in team red miner settings. You might be able to get away with better efficiency as my PSU isnt good and probably isn’t actually providing the 12v but I have a new one on the way. I’m pulling 610w at the wall