vfio VS Single-GPU-Passthrough

I was mainly following this guide to setup the VM and scripts: https://gitlab.com/Karuri/vfio

I've successfully set up single GPU passthrough as per https://gitlab.com/Karuri/vfio and https://github.com/joeknock90/Single-GPU-Passthrough .

- and a few other things mentioned here and here,

great thanks for the tip! havnt checked the logs for libvirt as i thought it was an osx issue. im using manjaro host, and have originally installed the vm using this https://github.com/foxlet/macOS-Simple-KVM, and this for the passthrough https://gitlab.com/Karuri/vfio

I used to have an RX580-based single GPU passthrough setup. You likely need the vendor reset module, since hot-plugging Polaris 10 GPUs is a bit buggy and can cause GPU lockups. It might also be worth dumping your GPU bios and configuring libvirt to use it so you get a visible boot screen (tool). I think I mostly followed this guide, however it hasn't been updated in quite a while it seems.

If you're up to the task you can follow this guide to get it working.

vfio-single-gpu-passthrough (this is for single gpu system, I link it for the section regarding hugepages, isolation, pinning & cpu governor which massively improved performance. It use script to dynamically run these only when you start a vm, so you can still use all of your cores when you stop the vm)

Link to hook scripts: https://github.com/joeknock90/Single-GPU-Passthrough

Yes but that's a different kettle of fish. You then don't need two gpus, although if you do it can be useful. Start here and see how deep the rabbit hole goes https://github.com/joeknock90/Single-GPU-Passthrough

Im assuming you're on a single gpu setup and, as such are trying to do single gpu passthrough with a single nvidia card. so you'll likely have to run some libvirt hooks or smth to run scripts to unbind your gpu once you start your vm. In which case id suggest going through these two guides 1 2 which go over completely passing through your single gpu when you start your vm

Single GPU passthrough: https://github.com/joeknock90/Single-GPU-Passthrough

Out of curiosity, who is Muta, the video dude, or the guide writer, or someone else? From the GitHub it sounded like running the commands from the guide would create a qemu.d/[VM Name]/prepare/begin/start.sh . If I were to simply create those directories myself, and then script as described in the video, I should be good?

Thanks for the info. Some of these I recognize from a single gpu passthrough project on github that i used on my main rig... https://github.com/joeknock90/Single-GPU-Passthrough