Awhile back I ran into some difficulty getting qemu to run the virtual machine how I would have preferred. I'm going to go back and give it another shot, once I upgrade my video card. I'm just wondering if it is possible to share a video card with hardware pass through, if I drop down to a tty to launch Windows.
My experience lies within OpenSuse & Xen where I did need a second graphics card.
That said, it might be possible to only use one (will try myself once the new kernel is released), but so far it would mean to disconnect the card from the host and re-attach it to the guest. Meaning the host kernel wouldn't have access to the card anymore, hence no tty.
It would work using ssh to connect to the host. I don't know how well your distribution handles the absence of a video card, but generally it should work.
As far as I know you can't share it, it needs to be picked up by vfio or pci-stub during boot. Once the kernel module is loaded for the card I'm not sure if you can safely unload it. Your best bet is to have a second card or on-board graphics (that's how I do it).
I don't know if sharing the host's gpu resources with a guest is quite possible yet.
It may be a thing soon, though.
I thought it was an issue with virtualization technology needing advancement. I just didn't know if it was theoretically possible to do. How, I look at it is if I'm buying two video cards I'd probably SLI them. If I was to get 4 video cards I'd probably rather 4 way SLI than have two 2 way SLI setups for separate machines. Besides, I don't have enough PCI-E lanes to support that many video cards, plus an NVME drive and network card or a second NVME drive.