I know from past experience that there are many ways to botch an installation of CUDA drivers and Toolkits, so my hope is that there are simple guidelines for what not to do. I already found out that it’s not good to assume that stock Linux CUDA drivers have anything to do with drivers written by NVidia (that was a surprise)
Current laptop/workstation has 2080Ti, with fresh install of linux Mint (Ubuntu) which does access the screen, apparently by generic drivers. I need to swap in official NVidia CUDA drivers and utils and libraries without
Are there any recommended procedures to stay out of trouble and avoid conflicting packages?