If sys-gui-gpu can, then gpu_HVM can as well?

User Support
1

Hello all. After 4 days of reading and testing I had to write.

I have dual Intel/Nvidia notebook.
HDMI and DP are tied to Nvidia.
When in Optimus mode, I can successfully create both sys-gui-gpu and Standalone HVM, both with only Nvidia attached and both fully working.

But, with sys-gui-gpu, obviously by design I can get external monitor to work over HDMI or DP (only, but exactly what I need) without any extra steps, just by starting it after automated creation, while Standalone HVM will load on a notebook screen only, though perfectly working either with nouveau or Nvidia drivers.

In addition, while Nvidia Standalone HVM is up, external monitor reacts and just goes black, but xrandr in HVM doesn’t see any provider except dummy0 (expected I’d say).

So, to back to the subject, if sys-gui-gpu in this scenario can go to external monitor, what guys you did to it, so I could do it to Standalone HVM as well so it could go to external monitor too?

Thanks a lot for any idea.

2

Any idea?
I added guivm, lightdm and dummy-psu services to Standalone, but that didn’t help to pass it to HDMI/DP.

What is in sys-gui=gpu that succeeds to pass it to HDMI/DP?

3

I have created Windows 11 HVM and attached NVIDIA GPU and Audio to it, and after I installed NVIDIA drivers in it I got external monitor output in Optimus mode.

So why in Linux HVM I don’t get external output, maybe this triggers someone’s idea?

4

I have succeeded, both with F38 and Win11 Standalone HVMs.

I can’t allocate more than 2112MB of RAM to them, but that doesn’t seem to affect their performance, at least for my use-case. Patching kernel both ways (via patch and xen.xml) doesn’t help, so will continue to research this separate issue.

1 Like
5

Check this:
Contents/docs/customization/gaming-hvm.md at master · Qubes-Community/Contents · GitHub

6

Thanks. As I wrote, I tried it, both that patch as well as @neowutran’s info about patching xen.xml.

7

For the continuity, now it is possible