GPUs are now including 1-3 HW encoders. It would be really neat if a single instance of Fab could take advantage of this and run multiple jobs in parallel!