So I’ve been working on a system to run some dashboard displays. I’ve got a X99 system with ESXi 6.7 installed. I’ve got GTX 770s as the cards for passthrough to the VMs for output to the 4K TVs I’m working with.
Windows doesn’t seem to have a problem using both SVGA (The virtual GPU from VMware) and the GTX770 once I add:
hypervisor.cpuid.v0 = FALSE
pciPassthru0.msiEnabled = FALSE
to the VMX. Everything works as it should and the 770 takes over once windows boots (which honestly is the best I expect).
Ubuntu is an entirely different story. (Ubuntu 19.04 BTW). VM is configured identically, BIOS boot, PCI device passed through, however adding the two above flags causes it to fail to boot (both the installer and a VM that I added the flags to after install).
The Device is recognised by lspci but the NVIDIA control panel loads to a white screen (Which indicates a driver issue, but remains no matter what driver I try).
The strange thing about this, is that the greeter loads fine, and in fact it loads on the GPU’s connected display, but once I login the desktop loads on the SVGA output. I thought this might be an issue with gnome selecting the first gpu in the system to load the desktop on, but xrandr doesn’t see either GPU, so I’m not exactly sure what might be causing it. Going to try Fedora next.
Big Edit: Yea Fedora works for some reason ok then…
Would love to fix Ubuntu though.