This ist simply not true. Yes some people behind gnome were the only only ones to support egl streams. But even new versions of gnome wayland shell still can not use nvidia-drivers properly and lead to a crash. The screen gets blank and freezes the system. And still developed Xwayland is just a compromise. Genuine Wayland is the overall goal to reach.
Instead of supporting GBM like all the other main GPU manufacturers do, NVIDIA again needs to go its own way and sticks to egl streams. Finally after a long dispute NVIDA apologized to developers. They promised a solution and poor James Jones was put up for this task. Since 2016 he tries his best to find a solution with the developers and nobody knows when that Solution ever is going to produce good results. Still the status of the software is prototype and just meets some of the basic functiones needed. But even if some developers achieve something usable with Jones together, their workaround and extensions can not perform as good as sticking directly to GBM instead. We donât have time to patch up compromises and work arounds. This is like an offense to the developers. Everybody knows, that valuable time wasted for EGLstream is needed and actually is used for patching actual applications. The Situation is so absurd and paradox that even NVIDIA themselves use forced low feature nouveau drivers instead their own ones.
I has been nearly 4 years now la hawla wa la quwatta illa bi llah and still nothing realy usable came out. Come on NVIDA, cooperate with the developers please! Thank You
Here you can follow up the sad status quo of that odyssey. https://www youtube com/watch?v=HYa4UvVtMOE&t=20690s
As from what I understand, GBM is vendor agnostic inside Mesa. As far as Iâm aware, to support Mesa, Nvidia would need a open sourced kernel driver. So to fully use GBM, Nvidia would need a open source kernel driver.
Then according to that youtube video, Nvidia talks about the limitations of GBM.
Support Vendor-specific capabilities, including fully opaque capabilities
Support image usage âtransitionsâ
Nvidiaâs hardware / driver (maybe both?) needs opaque buffers and finalization. So they canât handle GBMâs flexibility in this regard. The way GBM functions would (and internally must do) cause quite a large performance regressions on Nvidia hardware. It has always been known Nvidiaâs hardwareâs performance is tied to their driver. Both are tuned highly for each other. A fine example is Nvidiaâs compression algorithm. Thereâs a reason why Nvidia hasnât needed large memory bandwidth. Their memory compression is the best in the industry and why say a card like a 980 Ti with only 384 bit interface kept up with a Fury X with on a 4096 bit HBM interface when it came to memory performance. Even though the Fury X had 512.0 GB/s of bandwidth vs the 336.6 GB/s on the 980 Ti.
Nvidia isnât just going their âown wayâ and not wanting to use open standards⊠They actually canât use GBM. That being said Nvidia really does need to figure out something⊠A certain major competitor to them has really increased their market share of Linux users due to a certain open source driver. And then another competitor on the horizon wanting to get into the discrete industry who too has open source drivers on Linux.
Iâve been a long standing semi-happy Nvidia user on Linux. All of this typed on Linux with a RTX card and an actual Gsync monitor. But they have really been falling behind these last few years⊠I really do hope Nvidia can do something to improve themselves within the community and tackle Wayland.
So the only way to fix the Nvidia limitations is to change hardware. Which is the best way to support wayland by graphical card industry? AMD or Intel? In my opinion Nvidia developers are negligent. Nvidia doesnât want to invest resources in Linux. So I will invest my money in other GPuâs industry.
well thatâs just a silly thing to state. nvidia DOES invest resources into linux. of all the reasons why, the biggest reason is cuda. as the industry runs off cuda and many who utilize cuda run linux in the professional environment. you the consumer might not think you take advantage of cuda but its the primaly reason why nvidia invests heavily into linux atm. you might think cuda doesnât apply to you but if you use hardware acceleration on linux with a nvidiaâs card, nvidiaâs hardware accelerated video encoder, nvdecode, utilizes cuda and is generally considered the best you can get on both linux and windows out of any vendor. even vdpau. one of the first major apiâs for video hardware acceleration was developed by nvidia. even amd (ati at the time) adopted it. long before intel came out with vaapi.
if nvidia didnât invest into linux they wouldnât be developing an api to replace gbm that all vendors can use because of the shortcomings of gbm. the problem with gbm is that it wasnât designed with nvidia in mind. for reasons that you can point at both parties.
there isnât a vendor who doesnât have hardware limitations. even amd uses memory compression these days which requires software. there isnât a manufacturer that doesnât combine their hardware and software together. amd rdna is more software bound than gcn was. its that each vendor is different. and each vendor wantâs to keep certain stuff behind closed doors. for example, it took amd about three years to bring open source support of freesync to linux (and the kicker is nvidia brought freesync support to their drivers on linux before amd brought open source support for it on linux), and they never brought their freesync over hdmi support as open source to linux as amd uses a proprietary protocol over hdmi to do it. and amd doesnât want to open that protocol yet. to the degree amd said âthey will help develop a open source solutionâ rather just opening up the one they have. so for example, amdâs hardware can do freesync over hdmi, but you need a closed source driver to do it for the moment.
in the terms of wayland right now if you want to use it the best routes is either intel or amd. amd provides better performance (since amd has high performance cards, like navi) but intel is slightly more stable as their open source driver stack is more mature (older). unless you use an older amd card like polaris which is generally considered very stable on the open source amdgpu driver. nvidia does support egl streams over wayland. its currently the only way to get gnome to work on wayland over nvidia. but its not all that great and performance isnât anything spectacular.