distributed rendering (on hardware) - realizeable ??

hi @ all.

Yes, i know this thesis might sound a big silly, but today i philosophized over a method of realizing “some kind of distributed” rendering, on more than once PC, connected over a 100/1000Mbit network.
But the problem would be, in my opinion, that i have no chance to use hardware-accelerated OpenGL to that ?!?!
I came to the point, where i thought that it could be possible to render different “parts” of the scene, saving the final color & depth buffer from the partially rendered images and then “blend” this different picture-pieces from different machines together, to one final image.
As you can guess, i’d like to use OpenGL for doing this and no selfwritten software-rasterizer.
The main problem while using OpenGL for a purpose like this is that i have no chance to get the calculated data back from the graphic-board to send them over the network to do other processings on it.

What are you thinking of this idea ? Has anyone heard of a website or project, where this idea has been translated to working code ?

[This message has been edited by DJSnow (edited 07-23-2003).]

The MPK (multipipe) library on SGIs is fully functionnal and does exactly that, except it’s a unique system with many graphics cards (so no need to use the network to synchronize). It can even composite an image in weird ways.


You can render sub parts of a scene graph on various computers and then cretae an IBR object from each depth buffer and then combine all IBR objects on a stand alone computer.


Take a look at Chromium. This open source project supports many different ways of “splitting” the rendering across a number of PC in a cluster. You’ll find information on the project on sourceforge.

Best regards,


@Ysaneye / ToolTechn / Niels… :

hey, thanks.
i will check the information, you mentioned