LookingGlass

ds are fine, but custom vendor cards often but not always have issues. My reference 6700 works fine, but a sapphire 6700 probably won’t. And the 7000 series is also fucky in a new way somehow. Gnif knows far more about this than me, and has basically thrown up his hands at how AMD doesn’t care. He’s made occasional posts about it on https://forum.level1techs.com/ Gnif is also responsible for Looking glass: https://github.com/gnif/LookingGlass When it comes to splitting a gpu into virtual ones with SR-IOV/MxGPU, that’s not really a thing with AMD. Whereas Nvidia will happily give you what you want if you shovel some money into their gaping maw, AMD won’t even give business customers the time if day if you aren’t worth billions. They very deliberately do not want the unwashed masses to use MxGPU. See: https://www.reddit.com/r/VFIO/comments/eqvn9z/amd_mxgpu_or_s... for a summary of the years of hopelessness on this functionality.

nitor or switch monitor inputs there is Looking Glass[2] which captures frames and shuffles them to the host over shared memory. It also handles input via spice. It's still a bit of work, but for me that's mostly on the Windows side rather than the passthrough stuff itself. [1]: https://github.com/duncanthrax/scream [2]: https://github.com/gnif/LookingGlass

based on some virtIO development that seems to be going on in macOS. Or maybe it's to virtualize iOS or iPadOS, who knows? Personally I use Proxmox, which is a Linux hypervisor based on QEMU / KVM but with other features and a web GUI for administration, to run Windows, macOS, and Linux simultaneously, as well as host containers for side projects and other self hosted services.

) - Arch Linux installation running KVM hypervisor Guest (A) - Primary Linux Install Guest (B) - Windows Install On boot I pass kernel arguments[3] to H to instruct it not to touch the graphics card. This gives me a completely headless boot and leaves my screen frozen on whatever the bootloader last had displayed. Once H boots up, it autostarts a VM for A, with the GPU passed in as well as the sound card and a USB controller. From A I can use my Linux install as normal. The interesting part is when I want to switch to Windows, I can ssh into H and kick off a script[4] that will tell A to suspend to disk (aka hibernate), which for all the GPU and other HW cares is the same as powering off, then to start B with all the same devices passed through. I can then do the same process to go back to my Linux guest. I haven't done this yet but I'm planning to put the swap/hibernate drives for both guests on an NVMe SSD so the suspend/resume should be super fast. The reason I have my main system as the Linux guest instead of just using the hypervisor is that loading and unloading the drivers for the GPU on the host is kinda wonky and has resulted in kernel panics for me (that I can't debug without a display). Additionally as far as I can tell there's no way to hotswap GPUs with display outputs in X so I'd have to restart all my desktop programs anyways. If you have an nVidia GPU you also might run into issues installing the drivers[5] inside the VM, and have to change some settings in the hypervisor to get around them. Let me know if you have any questions, I'm happy to help or provide more information. Ask here or see email in bio. [0] https://reddit.com/r/vfio [1] https://wiki.archlinux.org/index.php/PCI_passthrough_via_OVM... [2] https://github.com/gnif/LookingGlass [3] video=efifb:off I don't have a good source on this, different systems may need different args [4] this script is not yet written, I have been using SSH from my laptop until now, I've only had this set up for like a week [5] search error code 43, it's also mentioned in [1]