"We have no plans to support Wayland."

[deanjo]

There are much bigger ram consumers, like the nvidia 260 series driver - when you want to use vdpau on systems with low memory you need very much swap space currently, you could not gain that amount while just using wayland. nvidia should better optimize the new driver for ram usage, would be better for all...

Hmmm what do you consider "low memory" . I'm running XBMC on a 8200 IGP with 260 series drivers (260.24) and the largest amount of memory that is used is the shared ram for video that has 512 Meg allotted to it. Actual ram usage with vdpau is < 250 Megs for the entire system running on opensuse 11.3 x86-64.

I know there was a vdpau mem leak in the earlier 260 series but that has since been resolved.

[Kano]

I don't know if you need to call it mem leak, but with 256.xx driver i can use a 1 gb atom system with 128 mb for vga without swap and with 260.xx driver this is completely impossible. With 2 gb ram the thing would look different as you definitely must have more than 1 gb ram...

[fjwhittle]

Is there any reason in the architecture preventing Wayland from having network transparency in the future? I honestly don't know, I never needed network transparency and never will. Wayland is not feature complete as of now, so maybe? Fragmentation can be bad (look at what happened to Compiz some time ago) but choice can be great and Wayland may really push forward development of drivers in so far a bit overlooked areas.

IIRC, the creator of Wayland said that network transparency was not an aim of the display server and is really something application and/or toolkit developers should think about. Another possibility is that someone could create a virtual X11 stack to run on top of Wayland, which would be sort of like a toolkit with network ‘transparency,’ and very much useful for supporting Legacy applications.
Fragmentation is not necessarily an issue here because X11 and Wayland are different projects trying to achieve different, if overlapping, goals. X.org and Wayland are both free software, and there is much capacity to improve one based on improvements in the other. Especially where it is in relation to the graphics drivers. This would be more like KDE vs. Gnome — they both run on X, but they never started as the same project; As such, the only issues that can be are the ones that fools make. And look how well those two projects are getting along, now.

As for nVidia supporting KMS, well if they do that would be great, and they wouldn't even need to write a framebuffer console – one of the things about KMS is that the fbcon module is not hardware specific – if anything understood the kernel driver other than their own proprietary OpenGL stack in X11.

[deanjo]

I don't know if you need to call it mem leak, but with 256.xx driver i can use a 1 gb atom system with 128 mb for vga without swap and with 260.xx driver this is completely impossible. With 2 gb ram the thing would look different as you definitely must have more than 1 gb ram...

Yes my main HTPC has 2 GB of ram for the simple fact that to get vdpau to run stutter free you have to allocate at least 512 of it to vga for it to work properly (my BIOS's will not allow a 512 Meg allocation to the IGP until 2 GB is installed in the system). However, the second XBMC that I do have that has only 1 GB of ram and a 512 Meg on the 8800GT still runs smooth without swap. I'm thinking that your running more into an issue with not having enough ram allocated to video (IIRC 256 Meg is considered the bare minimum for vdpau).

[deanjo]

Mythtv seems to agree with me on the video memory allocation.

http://www.mythtv.org/wiki/VDPAU

Many playback issues can be resolved by making sure enough memory has been allocated to the video card, 256MB is a bare minimum requirement and 512MB is strongly recommended. For on-board video this is usually configured via the "GPU Window" BIOS option.

[deanjo]

Stephan Warren seems to say the same thing as well,

http://www.nvnews.net/vbulletin/showthread.php?t=129440

It's probably simplest to get a 512MB system. Various users with 256MB have reported success, but it depends a lot on what else is running on your system

[Kano]

It definitely works with with older driver to decode full hd content using mplayer and xbmc with the auto settings in bios and 1 gb ram. Just vlc is a no go - too slow.

[deanjo]

It definitely works with with older driver to decode full hd content using mplayer and xbmc with the auto settings in bios and 1 gb ram. Just vlc is a no go - too slow.

Well that could be attributed to using the VAAPI wrapper perhaps.

[Julius]

I think everyone should keep in mind why Nvidia did their Linux driver in the first place... usage in the commercial 3D rendering sector and big players like Disney using Linux in their production pipeline.
There is basically no interest regarding the support of home-users, and having stable "gaming" drivers for Linux was just a side effect, so to say.
That is why Nvidia will not support Optimus on Linux (only used in Laptop), and why Wayland support is unlikely (since it is a geared toward consumers... e.g. desktop Linux use).

[RobbieAB]

Lastly, X may not die soon, but without support from Ubuntu, I think it will cause the development of display servers for UNIX-like OSes to fragment.

Not wishing to flog a dead horse here, but I thought one of the long standing gripes with Ubuntu was the lack of code contributed back to parent projects... Will Xorg even notice if Ubuntu goes and plays in their own sandpit?