Something is bugged in the new version if you set a fixed gridsize, instead of just the intensity. The hashrate jumps every few minutes up or down. It doesn't matter if it's a single rig, or a multi-rig. All cards jumps up or down together in the same time.
2.1.20 was fine, 100% stable hashrate with intensity (sorry, never run with gridsize)
2.2.1 haven't tested yet, so dunno if that bug excist there too
Windows 10, latest updates. nVidia 419.67. Tested with 1080 Ti's, 2080 Ti's, 2070's.
Here are two screenshots from just one card. You can clearly see the jump from 3.5 MH/s to 4.3 MH/s, between it jumped from 4.3 back to 3.5 and 10 minutes later again from 3.5 to 4.3. The time between jumping isn't fixed, it can be few minutes, or 10-20 minutes, or anything else. But its jumping all the time on every rig.
Edit: just did a quick 1-hour-test on that rig. -i 16 gives a 100% stable hashrate of 3.3. Going back to -gs 41472 its jumping to 3.5 or 4.3. It doesn't matter which gridsize, its always jumping up or down. So the gridsize seems to be buggy.
Edit 2: Also checked the core clock between the jumps, just in case. But the core stays always in the same range, no matter if it jumps up or down.
Edit 3: Will also check the pool-hashrate later after few more hours running. Will it be the lower hashrates, or will it be the average between low and high...
Edit 4: Checked again the older 2.1.20 with the same gridsize which I got from 2.2.2 auto tune. There are no strickt jumps from (for example) 3.5 to 4.3 and 4.3 to 3.5. But a very smooth sinus-wave between (in this case) 3.7 to 4.0 to 3.7 to 4.0... Strange picture and dunno if this is supposed. Can you please check that? Pool is 2Miners if this matters. Thanks!