Announcement

Collapse
No announcement yet.

AMD Posts Patch Enabling Vega APU/GPU Support For Blender's HIP Backend

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

  • #11
    Originally posted by Paradigm Shifter View Post
    ROCm has a trinity of problems, all of which feed into each other - hardware support, software support, and performance.

    Performance will improve as people become more familiar with coding quirks, and the development stack matures. CUDA wasn't exactly earth shattering back in the early days either, it's just that the only competition was from CPUs... which in some very specific (small, highly parallelisable) scenarios couldn't compete, so CUDA looked amazing (and it was).

    "Heterogeneous computing" was the phrase thrown around a lot by AMD with the early APUs (I even have a book on OpenCL development co-authored by an AMD guy) but it really felt like an absolute joke and seemingly quite quickly got shuffled off into the corner wearing the Dunce Cap.

    What kills ROCm for me is hardware support. I really, really, really want to get away from my absolute dependence on CUDA, but I cannot in any way, shape or form justify a system which actually officially supports ROCm without some evidence that it will work well for us. That means working, (fairly) portable proof-of-concept code which doesn't require ritual sacrifice and the alignment of the stars to get the hardware and software to do what I can do with CUDA by slapping an nVidia GPU in a PCI-E slot, typing apt install cuda and rebooting.

    I really hope the recent announcements of all-AMD supercomputers will spur further broadening of the support umbrella for the cards which aren't either a) really old, b) impossible to buy or c) insanely freakin' expensive.

    Official support for the GPU in the mobile 5000/6000 series, or the 5xx0G chips, or (please?) 6800(XT/M)/6900XT cards would go a long way.

    But for now I'm resigned to continuing dependence on CUDA.
    A couple things -
    • navi21 is officially supported by ROCm already. Don't focus too much on the specific SKUs - navi21 is anything from 6800, 6800 xt, 6900xt pro W6800 and the x50 variants.
    • Blender and ProRender already support HIP on Windows on navi21. This patch will help support on vega as well.
    I don't get where all this outdated info comes from.

    Comment


    • #12
      Originally posted by someguyhere View Post

      A couple things -
      • navi21 is officially supported by ROCm already. Don't focus too much on the specific SKUs - navi21 is anything from 6800, 6800 xt, 6900xt pro W6800 and the x50 variants.
      • Blender and ProRender already support HIP on Windows on navi21. This patch will help support on vega as well.
      I don't get where all this outdated info comes from.
      You can also install rocm easily on some distros using 3-4 commands. Sure, it's not as simple as a single command, but you get more fine-grained control over the features you want to install. https://docs.amd.com/bundle/ROCm-Ins...Cm.html#d20e49

      As for performance, NVIDIA GPUs have higher compute TFLOPS than AMD. So you can't really expect a navi21 GPU to match a 3090 for example. The compute perf of navi21 is more in-line with RTX 3070.

      Comment


      • #13
        Originally posted by someguyhere View Post

        You can also install rocm easily on some distros using 3-4 commands. Sure, it's not as simple as a single command, but you get more fine-grained control over the features you want to install. https://docs.amd.com/bundle/ROCm-Ins...Cm.html#d20e49
        Installing it might be easy on a supported version. But packaging it is massive pain, with so many problems (that have been getting fixed one by one has to be said), that it takes a ton of time for maintainers to patch all of those. The effects that the only fully functional version (as far as I know) is in Arch's AUR, Debian's is almost there, on Fedora so far only OpenCL is working...
        So if you you want to install it on your current distro installing might be super hard or next to impossible. And Nvidia "just works", even with its flaws.
        To be fair, all the people involved are doing a great job, it just isn't there yet.

        Comment


        • #14
          Actually, installing ROCm and HIP it isn't even possible on RHEL9/FedoraAnything. Does anyone know what is up with the RedHat side of things? RHEL9 has been out for a bit and there is still NOTHING available.

          Comment


          • #15
            Originally posted by Porcelain Mouse View Post
            Actually, installing ROCm and HIP it isn't even possible on RHEL9/FedoraAnything. Does anyone know what is up with the RedHat side of things? RHEL9 has been out for a bit and there is still NOTHING available.
            https://repo.radeon.com/rocm/rhel9/
            I believe the 5.2 release will include initial RHEL 9 support - GA should be next week some time.
            Test signature

            Comment


            • #16
              Originally posted by JacekJagosz View Post
              I hope distro maintainers can backport this patch. But I wonder how many distros actually enabled HIP support, it might be only Arch AUR.
              Ubuntu/Redhat/SUSE are supported by AMD.

              Comment


              • #17
                Originally posted by Svyatko View Post

                Ubuntu/Redhat/SUSE are supported by AMD.
                I mean about the Blender build in the repositories. Like if you do `apt install blender`, is the hip support enabled, and if it is, will they backport this patch.
                I expect only AUR has a blender build that supports HIP, and was wondering if any other distribution as well.

                Comment


                • #18
                  Originally posted by loganj View Post
                  i wonder how fast will this be on APU compare to the host CPU
                  About +30% in speed for CPU+GPU compared to CPU-only with Blender 2.9x + OpenCL.
                  GPU-only is about 30% slower than CPU-only, i.e. useless in that case because of stuttering DE.
                  And no stutter for DE when using CPU-only path.
                  CPU for the cost of APU will be faster in Blender.

                  Comment


                  • #19
                    Originally posted by Spacefish View Post

                    You have to look at it like that: ROCm was made for GPU Compute customers, so it was initially limited to products which where meant for this market. Luckily they had the same Architecture or at least a similar one GFX9 (GCN) or Vega was seen in consumer cards (Radeon Vega 56/64) as well as in Compute cards (MI50/MI60).
                    In the next generation they have splitted the architecture into a Compute (CDNA) and a graphics architecture (RDNA). Although they share some similarities, they are different chips with different architectures, memory layouts and so on.

                    ROCm did only support CDNA (MI100/MI200 Aldeberan and so on), which are 10k$+ Cards without graphics output.

                    If you look closely in ROCm there are hand written machine code kernels for a lot of the compute functions, as the architectures differ, they did not do it for the consumer cards / RDNA, as there was no software using these + they are a lot slower than the CDNA for most compute tasks, so "compute" customers are not interested in them..
                    Hardware support for older generations is dropped pretty quickly, as the large datacenters always choose the cheaper options between power cost and buying new, more efficient, hardware. As soon as the older gen is no longer attractive in this metric, there is next to no customer for AMD for this functionality.

                    But i agree, they should add first class support for all consumer architectures on launch day, as this might not pay off directly, but platform adoption will sure increase sales of their CDNA offerings.. In the end you need someone who writes software for their GPUs / Accelerators, if a student canĀ“t get a foot into the door with the card he already has, they will probably wander off to the competition and later on in their job will choose the platform they already know.
                    I don't want to hear about difficulties that AMD developers have to overcome.
                    I want to buy product and use it. Right now Nvidia's products work in that case, AMD's - not.

                    Comment


                    • #20
                      Originally posted by JacekJagosz View Post

                      I mean about the Blender build in the repositories. Like if you do `apt install blender`, is the hip support enabled, and if it is, will they backport this patch.
                      I expect only AUR has a blender build that supports HIP, and was wondering if any other distribution as well.
                      openSUSE has troubles with packaging Blender: new version is only available for Tumbleweed https://software.opensuse.org/package/blender
                      And AMD ROCm is only available for Leap 15.3.
                      User has to download tar.xz from Blender's site.

                      Comment

                      Working...
                      X