Given pretty much every conversation on the topic of ML here advises people to use the proprietary NVidia ML tools with proprietary NVidia graphics drivers for

Given pretty much every conversation on the topic of ML here advises people to use the proprietary NVidia ML tools with proprietary NVidia graphics drivers for NVidia cards, I'm failing to see this as any better or worse.