It seems that Cryosparc2 is missing “CUDA device numbers” setting, which used to be in “Compute” section of refinement job setup. I found this option very useful when setting up queues with demanding jobs (for example refinements with large box size). When a node has multiple GPUs, cryosparc would allow two such jobs run in parallel causing a crash due to memory limitations. With the previous version, one way to go around this was to specify the same CUDA device for both jobs, which would allow the queue execute without issues. Is this setting hidden somehow in Cryospacr2? Otherwise, is there another way to limit a queue to one job at a time, without waiting with manually submitting the next job until the previous one is finished?
Peter