There are several areas in which the NVIDIA driver lacks feature parity between X11 and Wayland. These may be due to limitations of the driver itself, the Wayland protocol, or the specific Wayland compositor in use. Over time this list is expected to shorten as missing functionality is implemented both in the driver and in upstream components, but the following captures the situation as of the release of this driver version. Note that this assumes a compositor with reasonably complete support f...
That’s a weird read having in mind I had to move to Wayland because x11 had severe screen tearing. I would have guessed Wayland had better support.
Wayland has a bunch of features that are so new they aren’t in the stable distros yet.
Nvidia went from declaring they were never going to support Wayland to trying to force their own EGLStreams stuff on everybody to reluctantly accepting the standard that was developed without them and trying to make it work for their driver. They’re playing catchup and it’s entirely their own fault for refusing to cooperate with anybody.
They’re moving more towards open source drivers now, probably because the people buying billions of dollars worth of GPUs to use on Linux servers for AI training have had words with Nvidia on the subject.
Not with everything, Nvidia stood still on explicit sync, in that case it were the idiots at Freedesktop that were massively blocking Wayland’s progress, trying to force an inferior technology, which Nvidia did not want to implement.
Okay, but that’s still partially on Nvidia for refusing to participate. They could have argued for explicit sync early in Wayland’s development but they weren’t at the table at all, so they got stuck with the technology that was decided on without them and had to argue for changes much later.
And they started off arguing for EGLStreams, but it didn’t work well either. Explicit sync came later.
EGLStreams is not superior to GBM.
It is better, just not on NVIDIA GPUs.
if your work requires CUDA then nvidia is the only option
Where does CUDA come into play for screen tearing?