My main workstation runs Linux. It has a second GPU (NVIDIA RTX 2080 Super), USB 3.1 card, and an NVMe drive passed to a guest via PCIe passthrough.[1] I have a

My main workstation runs Linux. It has a second GPU (NVIDIA RTX 2080 Super), USB 3.1 card, and an NVMe drive passed to a guest via PCIe passthrough.[1] I have a 2x2 DisplayPort 1.4 KVM to drive my monitors with the host GPU on one side, and the guest GPU on the other side. The peripherals are connected to the host through any open USB port, and the guest through the PCIe add-in card. Audio is handled with Scream[2] mostly so I can get >65536Hz sample rate. (Really terrible things seem to happen if you try to boot a qemu guest w/ the emulated audio attached to pipewire-pulse when the DSP graph has a 96/192KHz sample rate. I've also had latency issues in the past w/ bonafide pulseaudio and the emulated audio card.) I do all my gaming and most of my browsing inside the Windows VM, which is bridged to my usual data VLAN. The linux host is where I do development work which lives on a separate experimental VLAN. Other than that I run a few LXC containers for various services needed for running the LAN. (DNS, mail, VPNs, etc.) - I just want that stuff logically separated so that they can either (a) be moved to my new workstation in 2024, or (b) if one breaks it can just be rebuilt from scratch without affecting the others. It's also nice because I can use whatever distro works best for that particular package. [1]: https://wiki.archlinux.org/title/PCI_passthrough_via_OVMF https://wiki.archlinux.org/title/PCI_passthrough_via_OVMF [2]: https://github.com/duncanthrax/scream https://github.com/duncanthrax/scream