Multi card acceleration
We have two redhat 7.3 machines with more than two heads. The newest has a Quadro 4 400NVS which seems to behave like two
cards. It has two PCI addresses.
I have configured XF86Config-4 to be similar
to the "BothTwinView" ServerLayout in the samples however screens 3 and 4 are part of another desktop (ie not :0.0).
I can enable xinerama which then makes all four screens part of the gnome desktop, however 3d acceleration only works on the first two screens. opengl windows in screens 3 and 4 are grey although everything else works.
Does nvidia really support more than two screens now? The linux page says they support 16 screens, but do they support opengl on them all?
Has anyone got more than two screens to work without enabling xfree86's xinerama?
Any help would be greatly appreciated. We are looking at purchasing a large number of these setups but we need them to work properly.
I dont have an answer for your question but I think you got further than
We ran into the same problem with the Quadro4 400NVS that Gnome only initialised two screens when we wanted all four.
The first two screens get the Gnome desktop and the other two get the
Xwindows crosshatchweave pattern only. I can move the mouse across
all four screens but it changes to just an "X" as it moves over the
I'd like to get to a situation where I can make a single gnome desktop out
all four screens.
I've attached a copy of my Config file just incase somebody can see something
glaring and wrong.
Appreciate any and all help,
Well i have a solution that i have been working on, although its really just a hack...
My simple trick is to start up the xserver WITHOUT xinerama...
then start a XGGI up ontop of that Xserver, with multiple targets (read more about libggi in the documentation on www.ggi-project.org) so basically it displays what you would expect the Xserver to display(un accelerate)...
Now for 3d acceleration... I wrote a opengl passthru library a while ago(http://r0gu3.codices.net/) that i have recently worked into an opengl splitter, which basically renders the complete context on all outputs, then shifts the matrices over for that specfic display...
Now mind you this library isn't complete (as in it only works with a handful of demos, and has several other limitations...) but I should have a release out in not too much longer ( this same problem affects all the opensource drivers as well)
Another thing that needs to be done as far as acceleration, is to port XGGI up to Xfree86 4.x... it is a completely seperate application and doesn't depend on your xserver BUT i want to add xfree86's xv extension and xinerama extension...
The only real (practical) downside to this is the fact that you technically are running 2 xservers...
I could post on this bulletin board when i have a release of this library if you would like :)
Oh and just a side note... if you can live without xinerama then the solution is simple... use a different window manager(such as twm) OR (if possible) start multiple instances on your different screens... (eg, :0.0 is you initial screen where you window manager is now, but :0.1 is the second screen, and you could run a command like this
just an example.. but then opengl acceleration will(should) work on multiple screens... )
Just for instance, at home, i have a gf2ultra(single head) and a gf2mx (dual head pci) and i can get opengl acceleration on all three monitors at once :) (although in two seperate contexts)
OpenGL is only supported on the first head when using Xinerama.
|All times are GMT -5. The time now is 05:00 AM.|
Powered by vBulletin® Version 3.7.1
Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Copyright ©1998 - 2014, nV News.