still getting the random CUDA crashing on one of my nvidia rigs with anything newer than 2.8c. PM doesnt restart properly and just hangs with a windows error.
manual restart is possible, but hashrates on certain cards gets severly reduced.
only proper fix is full reboot.
works perfectly and will run hundreds of hours on 2.8c without an issue.
Is there any difference in the behavior if you specify -nvnew 0 (i.e. using the old kernels) on PhoenixMiner 2.9e?
just had a chance to test this. it seemed better running -nvnew 0, however i did have an overnight crash, and one card seemed to have lost its OC, but at least the miner restarted itselt and didnt just hang in windows.
went back to 2.8c again.