I have a pile of “still officially supported” nvidia boards that stopped working with things like compositors and screensavers about 5 years ago.
At that point I switched to amd, because “works and is 90% the perf/$” wins over “fastest brick”.
Maybe someday, nvidia will be a reasonable option for casual users under Linux. (I “just” needed to fork xscreensaver and upstream patches to the desktop environment to keep using the old cards —- would have taken a week).
I know they’re still preferred for engineering workstations, where there’s a full time person to screw with nvidia dependency hell.
If you’re in the Bay Area and want to see what I mean, go to microcenter and try to get the floor model dgx-spark to run an inference request. That’s an nvidia linux distro running nvidia example code on nvidia hardware, and it just… doesn’t work.
This is much better than I would have guessed!
I have a pile of “still officially supported” nvidia boards that stopped working with things like compositors and screensavers about 5 years ago.
At that point I switched to amd, because “works and is 90% the perf/$” wins over “fastest brick”.
Maybe someday, nvidia will be a reasonable option for casual users under Linux. (I “just” needed to fork xscreensaver and upstream patches to the desktop environment to keep using the old cards —- would have taken a week).
I know they’re still preferred for engineering workstations, where there’s a full time person to screw with nvidia dependency hell.
If you’re in the Bay Area and want to see what I mean, go to microcenter and try to get the floor model dgx-spark to run an inference request. That’s an nvidia linux distro running nvidia example code on nvidia hardware, and it just… doesn’t work.