3080Ti core clock automaticly goes down and keeps on 975mhz

[kgdk]

I have a galax rtx 3080ti and I set the core clock and lock it at 1350mhz on hiveos, however, when I start the nbminer, the nbminer shows that the gpu core clock is 975mhz not the number I set. Because of this, the hash rate is only 81mh. I tried some methods to lock the core clock but finally the core clock was still on 975mhz. Does anyone have some ideas on it?

[mathelm]

This won’t help, but I have two EVGA 3080 TIs, one XC3 and one FTW3.
The XC3 with the normal sized cooler is power limited and will only do 80-82 mh, whereas the FTW3 with the massive cooler and no power limit did 87+mh in NBMiner.

I switched last nite to T-Rex and the XC3 (now with exact same timings as the FTW3) does 79-81+ mh, and the FTW3 does 89-91+

But one thing that did help the XC3 was upgrading the nvidia driver to 495.44. On 470 (when hive removed 495 a couple weeks ago), the XC3 was down to 73-74 mh, while none of my other cards seem affected. But now have both rigs on 495.44
From a shell “nvidia-driver-update 495.44” in case you need it like I did.

[Mozgala]

Try to set PL to max allowed per card and see if it helps, power drawn will be limited by locked core anyway. Got similar problem with one of my cards and it helped.

[kgdk]

So sad that setting PL to max allowed on every card did not help on my 3080ti. Anyway, thanks for your idea, dude. I am now trying to flash bios to fix it.

[woodruffe7]

I’m having the same problem with my MSI RTX 3080 ti. This was after renewing the thermal pads. After reading your story it makes me think that it may be a coincidence. It is happening on my Windows test rig too, so the problem is not specific to Hive.