Low hashrate 3060Ti LHR under WSL2

[stevekstevek]

I know this isn't the optimal setup for mining, but I'm trying to mine using a 3060 Ti LHR under WSL2.

My system /is/ set up properly to have CUDA GPU access under WSL, and other miners do work (for example, I get ~93 MH/s using Autolykos from Autolykos2_NV_Miner.

When I try to use kawpowminer, though, I seem to be getting a very low hashrate of just over 2 MH/s:

 m 15:42:35 kawpowminer 0:04 A1 2.07 Mh - cu0 2.07
 m 15:42:40 kawpowminer 0:04 A1 2.08 Mh - cu0 2.08
 m 15:42:45 kawpowminer 0:04 A1 2.11 Mh - cu0 2.11
 m 15:42:50 kawpowminer 0:04 A1 2.17 Mh - cu0 2.17
 m 15:42:55 kawpowminer 0:04 A1 2.12 Mh - cu0 2.12
 i 15:42:59 kawpowminer Job: 265edbd1… block 1954554 localhost [127.0.0.1:6060]
cu 15:42:59 cuda-0    Launching period 651518 ProgPow kernel
cu 15:42:59 cuda-0    Pre-compiled period 651519 CUDA ProgPow kernel for arch 8.6
 m 15:43:00 kawpowminer 0:05 A1 2.15 Mh - cu0 2.15
 m 15:43:05 kawpowminer 0:05 A1 2.13 Mh - cu0 2.13
 m 15:43:10 kawpowminer 0:05 A1 2.15 Mh - cu0 2.15
 m 15:43:15 kawpowminer 0:05 A1 2.17 Mh - cu0 2.17

The benchmark does a lot better:

~/m/kawpowminer$ ./build/kawpowminer/kawpowminer --benchmark 1


kawpowminer 1.2.4+commit.632f6ea0
Build: linux/release/gnu

No OpenCL platforms found
 i 15:33:35 kawpowminer Selected pool localhost:0
 i 15:33:35 kawpowminer Established connection to localhost:0
 i 15:33:35 kawpowminer Spinning up miners...
cu 15:33:35 cuda-0    Using Pci Id : 2e:00.0 NVIDIA GeForce RTX 3060 Ti (Compute 8.6) Memory : 8.00 GB
 i 15:33:35 sim       Epoch : 0 Difficulty : 4.29 Gh
 i 15:33:35 sim       Job: 9676cf54… block 1 localhost:0
 i 15:33:36 sim       Using block 1, difficulty 1
cu 15:33:36 cuda-0    Generating DAG + Light : 1.02 GB
cu 15:33:38 cuda-0    Generated DAG + Light in 2158 ms. 6.98 GB left.
cu 15:33:39 cuda-0    Pre-compiled period 0 CUDA ProgPow kernel for arch 8.6
cu 15:33:39 cuda-0    Launching period 0 ProgPow kernel
cu 15:33:39 cuda-0    Pre-compiled period 1 CUDA ProgPow kernel for arch 8.6
 m 15:33:40 kawpowminer 0:00 A0 0.00 h - cu0 0.00
 m 15:33:45 kawpowminer 0:00 A0 6.34 Mh - cu0 6.34
 m 15:33:50 kawpowminer 0:00 A0 22.82 Mh - cu0 22.82
 m 15:33:55 kawpowminer 0:00 A0 22.23 Mh - cu0 22.23
h m 15:34:00 kawpowminer 0:00 A0 22.26 Mh - cu0 22.26
^[[B m 15:34:05 kawpowminer 0:00 A0 22.30 Mh - cu0 22.30
 m 15:34:10 kawpowminer 0:00 A0 21.93 Mh - cu0 21.93
 m 15:34:15 kawpowminer 0:00 A0 22.07 Mh - cu0 22.07
 m 15:34:20 kawpowminer 0:00 A0 22.28 Mh - cu0 22.28
 m 15:34:25 kawpowminer 0:00 A0 22.54 Mh - cu0 22.54
 m 15:34:30 kawpowminer 0:00 A0 22.04 Mh - cu0 22.04
 m 15:34:35 kawpowminer 0:01 A0 22.11 Mh - cu0 22.11
 m 15:34:40 kawpowminer 0:01 A0 22.27 Mh - cu0 22.27
 m 15:34:45 kawpowminer 0:01 A0 22.65 Mh - cu0 22.65
 m 15:34:50 kawpowminer 0:01 A0 21.99 Mh - cu0 21.99

I know kawpow and autolykos hashrates shouldn't be the same, but this seems like something is very wrong -- and not just due to the overhead of more expensive calls across WSL->Windows->GPU.

It is loading up my GPU (it shows 99% GPU core load, and ~100W consumption), so if there's an anomaly it's not as crazy as trying to mine on CPU..

1. Are the results I got close to what I ought to be seeing natively?
2. If this is slower than expected, any pointers about where to look to diagnose?