Announcement

Collapse
No announcement yet.

Making Use Of Intel vGPU Support On Linux 4.16 & QEMU 2.12

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Making Use Of Intel vGPU Support On Linux 4.16 & QEMU 2.12

    Phoronix: Making Use Of Intel vGPU Support On Linux 4.16 & QEMU 2.12

    As of the Linux 4.16 kernel that was released one week ago, the kernel-side bits are in place for Intel Virtual GPU support and in user-space the upcoming QEMU 2.12 has the necessary code for the GTK and SPICE code-paths...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    I am getting more requests for vGPU support in the development space. Actually so they can develop against non-Intel GPU's primarily.

    Comment


    • #3
      Does anyone know if this will work on Windows guests? (e.g. so that I can have full DirectX hardware acceleration on Windows in a VM)

      Comment


      • #4
        I'm so disappointed at how Intel just suddenly dropped gvt support on Haswell, after years of supporting it.. And I was already pissed finding out about Vt-d missing ACS support for pci-e root ports.

        Comment


        • #5
          Originally posted by fld- View Post
          I'm so disappointed at how Intel just suddenly dropped gvt support on Haswell, after years of supporting it.. And I was already pissed finding out about Vt-d missing ACS support for pci-e root ports.
          When did they drop it? Broadwell and higher is required for the splitting/sharing of the iGPU resources to multiple VMs, the Haswell and few generations lower was full passthrough of the iGPU which meant the host needed to use something else or be headless.

          Comment


          • #6
            Haswell was supported year or two ago. They really dropped its support prior merge to mainline kernel??

            Comment


            • #7
              Originally posted by edwaleni View Post
              I am getting more requests for vGPU support in the development space. Actually so they can develop against non-Intel GPU's primarily.
              Can you explain more? what does "develop against non-intel GPU" mean here?

              Comment


              • #8
                up until 2016 Q1, the gvt commits/slides would frequently mention Haswell support:
                "Intel GVT-g is a full GPU virtualization solution with mediated pass-through, \
                starting from 4th generation Intel Core(TM) processors with Intel Graphics \ processors."

                but something changed in 2016 Q2:
                " Intel GVT-g for KVM (a.k.a. KVMGT) is a full GPU virtualization solution with \ mediated pass-through, starting from 5th generation Intel Core(TM) processors with \ Intel Graphics processors."

                IOMMU/Vt-d/gvt support were the primary motivators why I choose to get a Haswell in 2014 in the first place.

                some random sauces:
                https://lists.01.org/pipermail/igvt-...ry/001075.html
                https://www.redhat.com/archives/vfio.../msg00029.html
                https://www.reddit.com/r/archlinux/c...kvmgt_working/
                https://marc.info/?l=linux-kernel&w=...t+announce&q=b

                Comment


                • #9
                  Meanwhile on amdgpu, the opengl still doesn`t work with QEMU. Intel`s GPU is pathetic but their driver support is there. If they made a half decent discrete GPU (say 1/4 the speed of a Vega) I'd buy it right away. Even if it was $500. Rock-solid open driver support is required. I get that it`s difficult, but AMD has got to get their act together. Even their amdgpu-pro driver doesn't work with kernels more recent than 4.9???

                  Comment


                  • #10
                    Originally posted by starshipeleven View Post
                    Can you explain more? what does "develop against non-intel GPU" mean here?
                    Their dev hardware contain Intel graphics. They want to program using some types of CUDA libraries. I suggested either using vGPU's in a remote setting or replacing their hardware with discrete. I also suggested using OpenCL to reduce the vendor tie in with CUDA. That is what I was referring to.

                    Comment

                    Working...
                    X