VDI

SR-IOV

https://wiki.archlinux.org/title/intel_graphics

grep '' /sys/kernel/debug/dri/0/gt/uc/*uc_info
dmesg | grep guc

Kubernetes cluster : https://github.com/intel/kubevirt-gfx-sriov

Ubuntu 22.04 LTS host: https://www.intel.com/content/www/us/en/developer/articles/reference-implementation/graphics-sr-iov-enablement-toolkit.html

https://github.com/strongtz/i915-sriov-dkms | PVE

for linux 6.1 ~ linux 6.3: https://github.com/strongtz/i915-sriov-dkms/blob/master/README.md

VirGL

https://pve.proxmox.com/pve-docs/chapter-qm.html#qm_display
https://docs.mesa3d.org/drivers/virgl.html

GPU assignment

The state of GPU assignment in QEMU/KVM: https://www.linux-kvm.org/images/b/b3/01x09b-VFIOandYou-small.pdf

https://wiki.debian.org/VGAPassthrough

https://wiki.gentoo.org/wiki/GPU_passthrough_with_libvirt_qemu_kvm

https://pve.proxmox.com/wiki/Pci_passthrough

Error handling

Windows BSOD: system thread exception not handled
echo "options kvm ignore_msrs=1" > /etc/modprobe.d/kvm.conf
Linux tty: Radeon kernel modesetting for r600 or later requires firmware-amd-graphics
apt-get install firmware-amd-graphics

virtio-gpu

https://www.linux-kvm.org/images/0/09/Qemu-gfx-2016.pdf

https://github.com/qemu/qemu/blob/master/hw/display/virtio-gpu.c

ARMv8/arm-64: use virtio-gpu because the legacy VGA framebuffer is very troublesome on aarch64

https://docs.opennebula.org/5.4/deployment/open_cloud_host_setup/pci_passthrough.html#loading-vfio-driver-in-initrd

Hardware

NVIDIA Certified

http://www.nvidia.com/object/grid-certified-servers.html GRID vGPU is a licensed feature on Tesla M6, Tesla M10, and Tesla M60.

VDI Solutions - Nvidia

http://www.nvidia.com/object/xendesktop-vgpu.html (Right bottom section: "Partner Solutions")

## Virtualized Application Solutions (Shared GPU)
Citrix XenApp 6.5 with OpenGL Add-on and Citrix XenDesktop 7 Hosted-Shared delivery

## Virtual Desktop Solutions (Shared GPU)
XenDesktop FP1 with NVIDIA GRID™ vGPU™1
Microsoft RemoteFX in Windows Server 2012
VMware Horizon View 5.2 with vSGA2

## Virtual Remote Workstation Solutions (Dedicated GPU)
Citrix XenDesktop 5.6
Citrix XenDesktop 7 VDI delivery
VMware View 5.3 with vDGA

AMD

AMD MxGPU is the world’s first hardware-based virtualized GPU solution, is built on industry standard SR-IOV (Single-Root I/O Virtualization) technology. Supprot VMWARE & Citrix.

Intel® Graphics Virtualization Technology

https://01.org/sites/default/files/users/u16902/graphics_virtualization_update_figure_1.png https://01.org/sites/default/files/users/u16902/graphics_virtualization_update_figure_2.png https://01.org/sites/default/files/users/u16902/graphics_virtualization_update_figure_3.png

–d (Intel® GVT –d): vDGA: virtual dedicated graphics acceleration (one VM to one physical GPU)
–s (Intel® GVT -s): vSGA: virtual shared graphics acceleration (multiple VMs to one physical GPU)
–g (Intel® GVT -g): vGPU: virtual graphics processing unit (multiple VMs to one physical GPU)

Visit Intel GVT-g page for more.

Optional Firmware

https://01.org/linuxgraphics/downloads/firmware

GuC：designed to perform graphics workload scheduling on the various graphics parallel engines. DMC: low-power idle states. It provides capability to save and restore display registers across these low-power states independently from the OS/Kernel. HUC: offload some of the media functions from the CPU to GPU.

ls /lib/firmware/i915/ -l

Citrix

XenServer

https://docs.citrix.com/content/dam/docs/en-us/xenserver/current-release/downloads/xenserver-configuring-graphics.pdf Graphics Virtualization is available for XenServer Enterprise Edition customers, or those who have access to XenServer through their XenApp/XenDesktop entitlement.