Ah, thanks for the tip! From what I can gather most of my services can run just as well with a minor tweak for GPU CUDA usage. For my case, it might not be a problem but using your method would no doubt be huge for Pi-based servers