Will running Pytorch CUDA 11.8 on CUDA 12.0 GPUs negatively affect performance?

Hi All,

This is a probably stupid question, but if I’ve installed the latest stable pytorch 2.0 with CUDA 11.8, but I’m running them on GPUs that have CUDA 12.0, will this negatively affect performance at all? I’ve installed via conda for reference, but I’m just wondering how this will affect performance, if at all?

Thanks!

Your locally installed CUDA toolkit won’t be used unless you build PyTorch from source or a custom CUDA extension, since the PyTorch binaries (pip wheels and conda binaries) ship with their own CUDA dependencies.

1 Like