A few years ago I had a PC with i5-4670 with iGPU (HD4600) and Nvidia GTX970 dGPU. My current setup is:
- Gigabyte X570 AORUS Elite
- Ryzen 7 5800X
- Nvidia GT210
- Nvidia GTX970
With my old setup I used to be able to leave my monitors plugged in the iGPU and use either the iGPU (when my GTX970 used the vfio driver for pass through to the VM i.e. ignored by Linux) or the GTX970 (with the Nvidia proprietary driver (nvidia) loaded i.e. active in Linux) for rendering and everything got piped through the iGPU to be displayed on the monitors (when GTX970 was passed to the VM I had to have a monitor plugged in directly to it to get output from it of course but that is not important in this case). I’m not sure whatever I used PRIME or Reverse PRIME or Render offload. I just had the mesa driver installed for the Intel iGPU and the Nvidia proprietary driver (nvidia) for the GTX 970 and when the Nvidia card was set to use the nvidia driver it just got piped automatically through the Intel iGPU. I think at some point I had something like this in my xorg.conf (along with other stuff generated by nvidia-xconfig):
Section "Device"
Identifier "intel"
Driver "modesetting"
BusID "PCI:0:2:0"
EndSection
so I guess I used Reverse PRIME? (Later it worked the same way without xorg.conf file)
My question is: Is this kind of setup possible with 2 dGPUs? Nvidia GTX970 will render and the other GPU will have the monitor plugged in. (This has to happen automatically without things like prime-run / bumblebee )
I know it’s not possible with 2 Nvidia GPUs but is it possible with Nvidia + AMD and Nvidia + Intel Arc?
Also I’m considering replacing my GTX970 with an AMD GPU in the near future so is this setup possible with 2 AMD GPU (if I get a low end AMD GPU now for AMD+Nvidia setup) or should I get an Intel Arc so that later I end up with AMD+Intel? (this question assumes that the answer to the first is yes)