On a Windows machine i am using hwmon to check my amd gpu memory overclocking. If there are memory errors shown on hwmon then i reduce memory clock until there are 0 or near zero errors. I know that gpu can still work even if lot of memory errors are prodoced and there are no other signs than hwmon indication. Tho I never do that on windows rigs. how to do mem overclock on linux? should i push cards as far as they dont reproduce bad shares on miners?
I have swaped from windows to linux, i have put all cards to 2050 mem clock. But i know that 2 of the card had errors on windows if pushed above 2000. but from linux side everything looks good! Can i leave it that way? Or are there some ways to check those errors? I am on hiveos.
With HiveOs probably you can squeeze more from your cards. Linux handles your cards much better than windows albeit it uses 5% more power in term of TDP power consumption. I have the only Linux rig with Rx 570 8Gb cards Sapphire Nitro+ and I have pushed the memory clock to the maximum at 2250 and everything is working well so far.
