GPU configuration with "env = WLR_DRM_DEVICES,/dev/dri/cardN" not working as expected?

[ssvx]

In reference to https://wiki.hyprland.org/hyprland-wiki/pages/Configuring/Multi-GPU/

My cards from lspci | grep -E 'VGA|3D' are

• 17:00.0 VGA compatible controller: NVIDIA Corporation GA102 [GeForce RTX 3090] (rev a1)
• 67:00.0 VGA compatible controller: Intel Corporation DG2 [Arc A770] (rev 08)

And from ls -l /dev/dri/by-path i get pci-0000:67:00.0-card -> ../card0, so i've set env = WLR_DRM_DEVICES,/dev/dri/card0 in my hyprland.conf, which should refer to my A770.

But even when i unload all nvidia kernel mods with sudo rmmod nvidia_drm nvidia_modeset nvidia_uvm nvidia before running Hyprland it seems to automatically load at least the nvidia driver and also seems to be running on that:

lsof -n -w -t /dev/nvidia*

5771

ps -p 5771

PID TTY TIME CMD
5771 tty1 00:00:45 Hyprland

Also when running kitty, Xwayland or chrome-based browser (Vivaldi) they all seem to jump onto the nvidia train.

Problem is that i want to attach/detach my 2nd GPU (nvidia) to be used on demand either as 3d accelerator for certain apps/games in Hyprland natively and wine or use the vfio drivers to pass the card on to qemu/kvm and use it as dedicated GPU for a virtual machine.

I've been told i'd need to blacklist nvidia drivers and reboot to use the GPU with vfio and then remove it from the blacklist and reboot again in order to use it within the host machine again. I find this very strange and cumbersome and would much prefer the GPU not being used as long as i configure Hyprland to use a different GPU and don't explicitly tell anything to load my nvidia kernel mods..?

Also: is there a way to check which GPU Hyprland thinks it's using to confirm the env = WLR_DRM_DEVICES,/dev/dri/card0 does anything at all?