No announcement yet.

Graphics Card Virtualization Levels & Status ?

  • Filter
  • Time
  • Show
Clear All
new posts

  • Graphics Card Virtualization Levels & Status ?

    I am trying to figure out what all levels of Virtualization can be achieved with Modern Graphics Cards
    # Virtualization Scenario Mobo/CPU Requirements Software
    • Complete GPU PCI pass-through to the Guest VM
    Intel VTd/AMD IOMMU kvm & similar Any Graphics card?
    • GPU & physical Video output ports partitioned for each VM (good for LEFTUVWFGs [Linux Enthusiasts Forced To Use Virtualized Windows For Gaming])
    • Setup
      • 1 Graphics Card exposes multiple virtual functions (virtual GPUs)
      • Each vGPU has its own PCI configuration space.
      • Pass-through is done at each vGPU level
      • Guest VM1 may use vGPU1 & Guest VM2 may use vGPU2.
    • Video Output
      • VM1 may output video to the DVI port
      • VM2 may output video to the HDMI port
      • VM3 may output video to the DiplayPort port
      • Bonus - Host can use vGPU0 & output video on say VGA port
    Intel VTd/AMD IOMMU Is there any combination of hardware that can do this as of now ?
    Are the current limitations in the silicon ? or in the firmware/drivers/kvm ?
    or is that so that it can be done purely in software by time-sharing the GPU, but not worth it ???
    • GPU Partitioned for each VM, but Video is streamed over a network over some remote desktop/terminal services solution
    • Setup
      • Similar to above
    • Video Output
      • No video output onto the physical ports(even if they are there)
      • Video is streamed over the network from the GPU hosting machine to the Client PC
      • Also called VDI (Virtualized Desktop Infrastructure)
    Intel VTd/AMD IOMMU AMD Firepro S7150/S7150X2
    4 GPUs that can do it all !! ??
    • What is the impact of Looking Glass, a KVM Frame Relay (KVMFR) ?
    • Is there a GPU+Software Combo that can do it all as of now ???!!!