Interesting results, in a nutshell it seems like Wayland/Xwayland performance on both nVidia and AMD wins slightly more than it loses. Once VRR is live in nVidia 545 series driver, for 3D games, Wayland is looking to deliver a great experience. Performance when Wine’s Wayland code is ready to mainline will be very interesting given that Xwayland needs will be negated at that point.
Wayland on GNOME (on Nobara 37) worked well enough for my use that I completely forgot it was using Wayland.
For KDE (on openSUSE Tumbleweed), that I’m currently using, I had to go back to Xorg but I have high hopes for Plasma 6.
I assumed that is because Wayland is at a point where devs are targeting it rather than X.Org now.
To the contrary, it’s more about the maturity and handling of XWayland itself. The developers are targeting AMD, specifically Mesa 23.x. The gains experienced by nVidia users is solely due to the optimizations made to the XWayland implementations. As far as “general use” for nVidia users, Wayland is blocked by the XWayland implicit sync standoff. nVidia refuses to support implicit sync because of “performance issues” while the X devs refuse to implement explicit sync because “Fuck you, nVidia!”.