VirtualGL with KVM PCI passthrough nvidia and vnc

[waltervargas]

Hi,

Host: Dell Precision 5510 with Dual Card Nvidia + Intel
Guest: Archlinux with the Nvidia card available via PCI passthrough.
+ VirtualGL
+ TigerVNC

Is possible to use the GPU of my Nvidia card for a VirtualGL setup inside the Guest in combination with tigervnc server?

[dcommander]

It should be, yes. VirtualGL doesn't really care about the driver-level details of the GPU. It just needs to have an X server set up on that GPU (we call that the "3D X server") so that it can create Pbuffers via the GLX API (Note that the 3D X server can be headless.) With PCI pass-through, the guest should behave just like a host from VirtualGL's point of view.

[waltervargas]

but the X server is not able to use the nvidia card, this is that kind of cards "hybrid graphics".

[dcommander]

I don't understand what purpose there is to having a GPU visible in the Un*x guest if it can't be used by an X server. When you say "hybrid graphics", I think of nVidia Optimus and similar solutions, but that's not the same as the virtualization environment you're describing.

At any rate, currently (and until/unless #10 is implemented), there is no way to use VirtualGL without a 3D X server. Full stop. So if you can't figure out how to run a 3D X server in the guest, or if such is not possible, then there is no way to run VirtualGL in the guest.