So I've been running various tests in HiveOS with different versions of the miner and different video drivers for the Nvidia cards (I messed around with the AMD drivers but keep on killing the OS, so I'm just going to leave that alone for now).

One thing I have noticed with the 1.42 Cuda115 release with the 495.46 Nvidia drivers (I'll check with 11.4 drivers later, but I don't recall seeing this), is that my 2060Super is only at  85-86% memory utilization (not the amount used but the time it's being used), while the 3090 is moving along at a solid 100%.

Here's a snippet of the test output:

root@local:~# nvidia-info
 
Wed Jan 19 12:08:16 EST 2022

=== GPU 0, 37:00.0 GeForce RTX 3090 24268 MB ===
  Bios 94.02.42.80.27, PCIE Link Gen 2, PCIE Link Width 1x
  S/N -, UUID GPU-e3c74e57-e377-7cda-4fd7-71caa2700bf2
  Power Limit:  Min 100.0 W, Default 420.0 W, Max 500.0 W, Current 500.0 W
  Frequency:  GFXCore 1590 MHz, MEMCore 11214 MHz
  Memory:  Total 24268 MB, Used 5025 MB, Free 19242 MB, Micron GDDR6X
  Utilization:  GPU 100 %, MEM 100 %, Throttle: None
  PSTATE P0, PWR 335.0 W, Temp 38 °C, Fan 89 %, RPM 2523, RPM 2536

=== GPU 1, 3D:00.0 GeForce RTX 2060 SUPER 7982 MB ===
  Bios 90.06.44.80.41, PCIE Link Gen 3, PCIE Link Width 4x
  S/N -, UUID GPU-24bf0276-6c6c-87d1-7922-8927a3d0b845
  Power Limit:  Min 125.0 W, Default 175.0 W, Max 215.0 W, Current 215.0 W
  Frequency:  GFXCore 1350 MHz, MEMCore 8062 MHz
  Memory:  Total 7982 MB, Used 4860 MB, Free 3122 MB, Samsung GDDR6
  Utilization:  GPU 100 %, MEM 86 %, Throttle: None
  PSTATE P0, PWR 97.9 W, Temp 35 °C, Fan 80 %, RPM 3040, RPM 3040

root@local:~# nvidia-smi stats -i 1 -d memUtil
1, memUtil , 1642612102256629, 86
1, memUtil , 1642612103259833, 86
1, memUtil , 1642612104263012, 85
1, memUtil , 1642612105266008, 85
1, memUtil , 1642612106270243, 85
1, memUtil , 1642612106270393, 85
1, memUtil , 1642612107273472, 86
1, memUtil , 1642612108276433, 86
1, memUtil , 1642612109280346, 86
1, memUtil , 1642612110283657, 86
1, memUtil , 1642612111286904, 85
1, memUtil , 1642612112289849, 86

root@local:~# nvidia-smi stats -i 0 -d memUtil

0, memUtil , 1642612118885645, 100
0, memUtil , 1642612119091821, 100
0, memUtil , 1642612119298340, 100
0, memUtil , 1642612119504453, 100
0, memUtil , 1642612119710978, 100
0, memUtil , 1642612119917647, 100
0, memUtil , 1642612120123717, 100
0, memUtil , 1642612120329955, 100
0, memUtil , 1642612120536173, 100
0, memUtil , 1642612120742582, 100
0, memUtil , 1642612120948902, 100
0, memUtil , 1642612121155221, 100
0, memUtil , 1642612121361320, 100
0, memUtil , 1642612121567466, 100
0, memUtil , 1642612121773858, 100

The 2060S is hashing at 40 MH/s but the client is reporting 44.75 (after 1 hour, so luck might be heavily influencing the result); do you know of any reason why utilization would be so low for the 2060S?