With TBM I had to slightly lower my mem overclock, it was crashing all the time. Even with lower mem overclock rig restarts about 20 times per day.
Try to add --dagintensity 1 and --kernel 1. (kernel 1 low power, kernel 8 high power)
The dev fee is mined in the first 10 minutes of mining, so if your rig is crashing 20 times, the devfee is higher than normal. Miner startup and Dag creation is also taking a while to create (up to 20 seconds), so a stable rig is important.
Ethermine reported hashrate in "ETH_WALLET1" is the full rig's speed not the 90% of the rig's full speed as it should (since the maintenance fee is 10%)
10% of the time is used to mine the maintenance wallet, so it will even out over time. 9 hours isn't enough. If the rig is crashing regulary, the factor might change.
The maintenance wallet is starting around 20mins into the mining and will run for a few seconds depending on the config, and then resume after another 20mins ..
If your rig is mining for 19 minutes and crash every time, 0 to the maintenance wallet..
I checked my memory stats on HiveOS and it seems that there is a some weird allocation happening in TBM 1.32 (maybe earlier too?).
We fixed a couple of leaks in the latest versions. Older versions have a leak in the dagger buffer when mining ETH+ZIL. 1.32 should be good though. You can restart and do some tests, report an issue to git if you find anything and we will fix it.
v1.33 will have an improved hashrate on the 3080ti (LHR) + 10MHASH stable. The 3080ti and 3060 LHR1 can run for 24 hours without triggering LHR, but I have problems with the 3070 LHR2.
[moderator's note: consecutive posts merged]