Announcement

Collapse
No announcement yet.

An End-User Has Made It Easier To Build ROCm & AMD GPU Machine Learning Software

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

  • #11
    Originally posted by Panix View Post
    Isn't that the norm or tradition with AMD and their gpus/software - just like the guy who 'hacked' CUDA for AMD gpus (i.e. Zluda)- AMD's support of their own software and gpus is atrocious - I guess the silver lining in all of this is FOSS that makes it easier for the end user - hooray! But, yeah, 'embarrassment' is the theme, here.
    People still do not seem to get that AMD has zero interest in GPGPU, which is why they keep slow-walking support.

    Under Lisa Su AMD made the decision to compete with "more cores" and this is on direct conflict with GPGPU.

    The only reason AMD even pretends to compete in that market segment is because of NVIDIA but I really believe that if AMD found a buyer for the old graphics division they would sell it in a heartbeat.

    Comment


    • #12
      Originally posted by timofonic View Post
      Also, include patches to upstream for consumer GPU support!
      The patches (for NAVI1) are upstream. They were already in place for 6.1, but weren't enabled by default. I heard they should be enabled in 6.2.

      Comment


      • #13
        Originally posted by skeevy420 View Post
        No, the bait is pointing out how the NVIDIA/CUDA EULA makes ZLUDA illegal software in countries that don't have strong reverse engineering protection laws.
        Such laws would effectively kill innovation, not to mention effectively nullify patent protections,

        It costs a lot of money to hire engineers and build machines to create new products, patents are what protect a company's investment from being stolen by a competitor that waits until the first company figures everything out and then comes in and just steals their work.

        Reverse engineering should be illegal until the patents expire.

        Comment


        • #14
          This article about the excellent efforts of Debian to build up a CI pipeline for ROCm is even more damning for AMD: https://news.ycombinator.com/item?id=37663194

          To be more direct, one thing we lack is funding. AMD has provided RDNA 2 and RDNA 3 GPUs for the Debian CI, but to fill out the rest of the architecture matrix I have been personally buying GPUs. That's been sufficient for covering most architectures, but we will need a sponsor if we are to acquire CDNA 2 and CDNA 3 hardware.

          Our goal is to cover every modern discrete AMD GPU architecture on the CI. At the moment, that would be Navi 33, Navi 32, Navi 31, Navi 24, Navi 23, Navi 22, Navi 21, Navi 14, Navi 12, Navi 10, Aldebaran, Arcturus, Vega 20, Vega 10, and (maybe) Polaris. I have been very successful at bringing the AMD GPU libraries to architectures that are not officially supported upstream. Unfortunately, I can't afford to keep buying systems out of my personal funds. I have personally spent ~7k USD on hardware for the CI and I have been offered reimbursement from the Debian project for my next ~5k USD in spending. That has given us a good foundation, but we could do more to improve hardware support if we had more funding available.
          That guy works for AMD. The fact that they aren't directly funding stuff like this is ABSURD. The money required wouldn't even be a rounding error on their P&L. Fix it Lisa Su!

          Comment


          • #15
            ZLUDA is a re-implementation of the CUDA API using ROCm, it doesn't disassemble, decompile, or reverse engineer any part of the Nvidia CUDA stack. It's similar to how Gallium-Nine works in that way. ZLUDA is not illegal, Nvidia worded their EULA in such a way to make it sound like it is, but it isn't. If it was, Nvidia would have sued the creator already rather than trying to scare people away from it instead.

            And that's not including the fact that EULA have been shown time and time again to not be legally enforceable in courts. That's another reason Nvidia put it in their EULA, because they can't do any real legal actions to ZLUDA or anybody using it.

            Comment


            • #16
              Originally posted by sophisticles View Post
              Such laws would effectively kill innovation, not to mention effectively nullify patent protections, It costs a lot of money to hire engineers and build machines to create new products, patents are what protect a company's investment from being stolen by a competitor that waits until the first company figures everything out and then comes in and just steals their work. Reverse engineering should be illegal until the patents expire.
              I'm sorry, but you sound like a businessman. Look at China. They ignore occidental patents and that made get tons of money. Patents are good for sharks. Patents kill innovation.

              Comment


              • #17
                Originally posted by sophisticles View Post

                People still do not seem to get that AMD has zero interest in GPGPU, which is why they keep slow-walking support.

                Under Lisa Su AMD made the decision to compete with "more cores" and this is on direct conflict with GPGPU.

                The only reason AMD even pretends to compete in that market segment is because of NVIDIA but I really believe that if AMD found a buyer for the old graphics division they would sell it in a heartbeat.
                While I disagree with you in patent sus. I agree they don't invest enough in GPGPU and seems they have zero or low interest on it.

                Comment


                • #18
                  Originally posted by pWe00Iri3e7Z9lHOX2Qx View Post
                  This article about the excellent efforts of Debian to build up a CI pipeline for ROCm is even more damning for AMD: https://news.ycombinator.com/item?id=37663194



                  That guy works for AMD. The fact that they aren't directly funding stuff like this is ABSURD. The money required wouldn't even be a rounding error on their P&L. Fix it Lisa Su!
                  Holy shit …..



                  Comment


                  • #19
                    They just enable the "officially" supported targets in the public binary builds.
                    It would be easy for AMD to enable all architectures / the projects like PyTorch ONNX and so on to enable them too.

                    Speculation: Marketing does not want to enable non officially supported products, as they fear the support requests.
                    The Engineers probably would happily enable everything in the official builds and just state that it´s untested / not officially supported.

                    Comment


                    • #20
                      Originally posted by pWe00Iri3e7Z9lHOX2Qx View Post
                      This article about the excellent efforts of Debian to build up a CI pipeline for ROCm is even more damning for AMD: https://news.ycombinator.com/item?id=37663194

                      That guy works for AMD. The fact that they aren't directly funding stuff like this is ABSURD. The money required wouldn't even be a rounding error on their P&L. Fix it Lisa Su!
                      Damn, when I say AMD doesn't care about their GPUs I meant it somewhat facetiously, but they genuinely don't care, do they?
                      I hate to say it but I'm glad I caved and just got an NVidia card, this is just stupid at this point.
                      And now NVidia is probably going to start making CPUs and eat AMD's lunch, purely due to optics, just because AMD is becoming notoriously outdated looking now that GPUs are basically powering the global economy with the current AI bubble. I think even Intel looks more forward-thinking than AMD does to the average investor at this point. This is gonna go in the books as a huge blunder and missed opportunity, like how Microsoft ignored smartphones until it was too late.

                      Comment

                      Working...
                      X