To read all comments associated with this story, please click here.
It's perfectly possible to run opengl with an exported X display as long as the driver supports it. As far as I'm aware nvidia & intel drivers already support this config but performance won't be great though
What I'd like to see is the server generating a realtime h.264 stream - anti-aliasing 3d stuff would essentially be done for free by the codec, with a decent quad-core chip or gfx card capable of realtime h264.
3D games are really low-latency realtime case: you need to deliver a frame between two vblanks, which is a problem over the network (network latency, en/decoding, buffering frames, etc.), especially if you spend time on en/decoding (also many codecs have frame difference stuff which may require more frames present to encode).
Edited 2007-09-11 08:48
Will these specification make it possible to write a driver so I can put a 3D card in my headless server, have all OpenGL calls redirected and computed on this server and the results returned over the network?
You know how some people say that because X is 30 year old technology, we should dump it and start again?
Well, what you're asking for is 30 year old extensible technology (X), and the not particularly new GLX extension (GLX, not Xgl, which is something different).
Next time you hear someone say we should dump X, tell them to hollow out the bone in their head to make room for a brain. :-)
(I think I misunderstood your 'server' to mean 'server' in the X sense. Rather than do the 3D rendering on the application server, do it on a thin client with cheap 3D hardware on board. You're still running your actual apps on the application server, and getting beautifully rendered stuff on the display. Dave Richards from Largo is doing precisely this => check out his blog linked from Planet GNOME.)
Edited 2007-09-11 07:32
Maybe I wasn't too clear in my question. The last possibility you mentioned as in the GLX extension is one option. You run your applications on the (headless) application server and display the output on your client having a massive 3D card using the locally DRI/OpenGL accelerated X server.
The other possibility is to have a 3D card in your (headless) applications server, do all OpenGL computations there and return the results somehow to your relatively weak client device with some older 3D card unsuitable to playing modern games with the local DRI/OpenGL accelerated X server.
For instance I have a Dell C600 laptop with an Ati Rage 128 Pro graphics card and it works fine for normal things. But when I want to play newer games I would have to buy a new laptop since I can't upgrade the graphics on this machine.
My suggestion is if I would put a modern AGP 3D card in my server (a repurposed old desktop machine) and run the 3D games and applications there and only display the output on my laptop in an efficient way, would it work so I wouldn't have to buy a newer laptop just because of the old graphics card.
Edit: I am probably thinking of what the Fusion project is going to with the integration of CPU and GPU, but that's still a few years away.
Edited 2007-09-11 09:51 UTC





Member since:
2007-01-27
Will these specification make it possible to write a driver so I can put a 3D card in my headless server, have all OpenGL calls redirected and computed on this server and the results returned over the network?
I am using VNC to connect to my server and everything is running there but I can't run 3D stuff this way. It would be great if I could run 3D applications and games on the server and transparently display them on my laptop in an efficient way!
The only thing is that this protocol probably wouldn't be called VNC (Virtual Network Computing) anymore but ONC (OpenGL Networking Computing).