Some GPUs go to 0 MH/s at DAG Epoch

[wtu-fl]

This has been a issue with the last few releases of Nanominer 5.x. I am running all AMD GPUs mining ETH in Windows 10 Pro on several rigs, mostly RX5700XT and RX6800 GPUs, and use Afterburner to preset all GPU parms and keep Afterburner running to control fan speeds. All GPUs always initialize fine when launching Nanominer. But when a DAG Epoch comes around, a few GPUs don't "survive" and go to 0.00 MH/s. I normally have dagSer=true, but it doesn't seem to make any difference. It is not always the same GPUs on each rig, either. Symptoms include the DAG generation for a GPU taking 5 times as longer than normal (for example, an RX5700XT might normally take 7.5 seconds to load DAG, but when this happens it might be 35 to 60 seconds.) Sometimes, Afterburner shows that the offending GPUs have lost their voltage and clock settings. To fix, I bring down Nanominer, use Afterburner to manually load the proper voltage and clock settings into each GPU, then restart Nanominer. So... is Nanominer inadvertently resetting GPUs or changing their settings on a DAG update?

[wtu-fl]

When this happens, it is almost always a RX5600XT or RX5700XT. It hardly ever happens on an RX6800.