Hi John, On Fri, Oct 8, 2010 at 5:40 PM, John Kelso <ke...@nist.gov> wrote: > Our next step will be to try mosaic mode as suggested by both Wojtek and > some folks from Nvidia. We'll create one big window that covers all four > cards and open 4 viewports on it, one viewport per card, and see what the > performance is.
This approach will force the OSG to run with a single graphics thread, and end up throwing 4 times as much data at this graphics context. On the driver side this graphics context will also have to have either be a proxy context that has four contexts underneath, or a context that is duplicated four times for each graphics card and cull on viewport if that's possible. Scalability wise this solution is crazy and smacks of hacks to get games to scale across multiple graphics cards when they are only written to render to one. This *really* isn't the way one should design software and drivers if you want maximum scalability. If this works then there is clearly some big issue with the graphics driver not handling a single process with multiple threads and multiple contexts. We can't rule out the OSG not managing things perfectly, but it's written to abides by all the right principles for getting things to scale. I'm intrigued to see how things pan out. Robert. _______________________________________________ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org