Well like I say it's crashing after 5 minutes and I can't use the ROCm kernel anyway (or amdgpu-pro driver either because I'm using Debian now but apparently you don't need that).
When you say it's 0.5Mh/s below Windows do you mean with or without the dag fix?
even with DAG fix the hashrates I get is 0.4-0.5 lower on same card same clocks on win 10.
Just to add to this now that I got it working and have some data...I've been testing 1 of my cards in Linux and it gets 28.5 (reported by Claymore's miner) using practically the same settings (-10MHz on the memory because 2060 seemed unstable), as opposed to 29.5 reported by Claymore in Windows.
GPU # 1 in this screenie:
https://i.imgur.com/pyOzP3i.pngHowever, the actual effective 24hr avg hash rate on the pool side (AFTER losses from stale shares and Claymore's fee) is consistently slightly higher than the
reported hash rate in Windows. Meanwhile the actual effective hash rate of the Windows cards is 5% or more
below the reported hash rate.
https://i.imgur.com/pB7tvSo.jpgThat said, Windows does actually seem more stable. I mined for over a month with no signs of stopping until I shut down the rig on purpose to do some maintenance. In Linux it crashes every once in a while even with a slightly lower mem clock and slightly higher power use and temps (so I'm assuming core voltage is slightly higher but I have no way to check for sure). Also no way to monitor GPU memory errors so I'm not sure if the crashes are due to memory clocks/timings or a lack of core voltage.