Luis Chamberlain sent out the modules changes today for the Linux 6.6 merge window. Most notable with the modules update is a change that better builds up the defenses against NVIDIA’s proprietary kernel driver from using GPL-only symbols. Or in other words, bits that only true open-source drivers should be utilizing and not proprietary kernel drivers like NVIDIA’s default Linux driver in respecting the original kernel code author’s intent.
Back in 2020 when the original defense was added, NVIDIA recommended avoiding the Linux 5.9 for the time being. They ended up having a supported driver several weeks later. It will be interesting to see this time how long Linux 6.6+ thwarts their kernel driver.
And the relation between graphical drivers and ML is?..
GPUs are the primary hardware used for ML workloads…
Yes, but you don’t need graphical drivers to do ML on a GPU, you just need the part of the drivers that lets you do ML. Accelerated graphics are almost completely unrelated. Nvidia can stop offering graphics drivers while still ofering ML drivers and still make a very good living.
Even if they continue offering graphics drivers they don’t have to offer them for free. Their main clients are people who do professional graphical and video editing, who can drop hefty sums on driver licensing because they already pay a lot for the hardware and support. Gamers are a tiny amount of their revenue, and over 90% of that tiny revenue is Windows anyway.
At this point Nvidia can snap their fingers and discontinue all their support for the consumer market just like that and won’t even feel it. The only reason they bother with free Windows graphics is that Windows gamers still generate a non-negligible amount of revenue by buying the overpriced desktop cards, and the only reason they bother with Linux graphics drivers is because it’s free beta testing. The Linux desktop market is a ridiculously tiny population in terms of gamers, but it’s a sizable population in terms of QA.
Yes but this article is talking about the entire nvidia kernel driver… Why are you assuming this doesn’t apply to the parts necessary for ML?
Cheap vector operations.
Life’s too short to do ML on a CPU. In some cases literally. You can’t do any reasonable ML without a GPU. And you need drivers for that.