I have a mix of Nvidia Maxwell and Kepler GPUs on my main machine, and I’m having an issue with CUDA not being available after a sleep. When the machine wakes up, apps say that CUDA isn’t available, even though nvidia-smi seems normal. This didn’t happen in F35, but it’s happening in F36 at the moment. I’ve tried turning on persistence mode with nvidia-smi, but no dice.
Has anyone run into this? Thanks!