I think the problem here is not necessarily the AllowNVIDIAGPUScreens
option itself, but the fact that it requires a ServerLayout
section to specify it. There’s a known issue with the X server where it will prefer a server layout it finds in xorg.conf.d/*
over one you specify yourself in /etc/X11/xorg.conf
. So even though your layout named Layout0
has two X screens, the server is using the one named layout
that it’s finding in /usr/share/X11/xorg.conf.d/10-amdgpu-nvidia-drm-outputclass.conf
that has an implicit one-screen layout. This causes the server to create a GPU screen for the other Device
section you specified in xorg.conf
.
I think you’ll need to delete the ServerLayout
sections in /usr/share/X11/xorg.conf.d/*
before your custom layout will be used. Alternatively, you should be able to start the X server with the -layout Layout0
option to choose your layout explicitly.
A future driver release will enable AllowNVIDIAGPUScreens
by default on xserver 1.20.7 and newer so hopefully this problem will go away relatively soon.