I have a little problem with XMR stak.

I have two mostly similar rigs mining Monero, one running Windows 10, another SMOS. Each has 8 x RX580 Strix.

Some time ago I managed to find a spot on config for the Windows rig. It was two workers per GPU and managed to squeeze about 500 H/s per worker.

For a while I couldn't figure out how to modify AMD.txt in SMOS to get the same result, but today I managed to do that. I pushed the same thread config files in persistent memory and pointed the miner to them. The config files do their job but hashrate is about 100 H/s lower per worker, so about 200 H/s lower per GPU.

That's a significant difference and I can't figure out what could the reason be. Anyone with the same problem?

P.S. I had the "mmap failed" messages, followed the FAQ and fixed it. Didn't make a difference in hashrate.

P.P.S. Windows rig has 12 gigs of memory and runs from large fast SSD. SMOS rig has 4 gigs of memory and runs from the USB 2.0 flash drive.

P.P.P.S. I have several GPUs from the SMOS rig underclocked compared to others for stability reasons. That usually results in those GPUs showing less hashrate, but now all GPUs show hashrate within 1 H/s from each other. Which makes me suspect that there's a "bottleneck" somewhere capping all GPUs at about 80% of their full potential.