Were you using the same setup on the different systems, i.e. same OS, PyTorch version, CUDA, cudnn, etc. or what were the differences?
Did you see any hardware issues in the first system before and could you run a stress test on the GPU?
Were you using the same setup on the different systems, i.e. same OS, PyTorch version, CUDA, cudnn, etc. or what were the differences?
Did you see any hardware issues in the first system before and could you run a stress test on the GPU?