Virtual GPU

AMD

• Product: firepro s7150

Nvidia

• https://docs.nvidia.com/grid/latest/grid-vgpu-release-notes-generic-linux-kvm/index.html
• https://www.linux-kvm.org/images/5/59/02x03-Neo_Jia_and_Kirti_Wankhede-vGPU_on_KVM-A_VFIO_based_Framework.pdf
• https://www.redhat.com/archives/libvir-list/2016-August/msg00939.html

Unlock GPU

• https://github.com/DualCoder/vgpu_unlock

Validated Platforms

This release of NVIDIA vGPU software provides support for the following NVIDIA GPUs on Red Hat Enterprise Linux with KVM, running on validated server hardware platforms:

• Tesla M6
• Tesla M10
• Tesla M60
• Tesla P4
• Tesla P6
• Tesla P40
• Tesla P100
• Since 5.1: Tesla P100 12GB

For a list of validated server platforms, refer to NVIDIA GRID Certified Servers

QEMU/Guest graphics acceleration

QEMU/Guest graphics acceleration

Intel

• kvmgt mode: you can split a recent-ish intel igpu into 16 virtual gpus and assign them with full acceleration in the vms

• https://github.com/intel/gvt-linux/wiki/GVTg_Setup_Guide

Comments from Telegram group

The best option imho is virgl actually, because for one it is completely platform-agnostic for hosts and runs pretty performantly with remote usage available (not sure if trough virt-viewer yet).

mxgpu, intel kvmgt, whatever nvidia does or vga passtrough all require you to either use looking glass for forwarding or use the output of that card directly so remote is not possible at all unless you remote to the vm itself.

GPU passthrough with SPICE

https://jugulaire.github.io/GPU-passthrough-KVM-QEMU/

Passthrough

• https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/7/html/virtualization_deployment_and_administration_guide/sect-device-gpu

• https://wiki.archlinux.org/title/PCI_passthrough_via_OVMF