Announcement

Collapse
No announcement yet.

NVIDIA Developer Opens Feature Pull Request For Open-Source NVK Driver

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

  • NVIDIA Developer Opens Feature Pull Request For Open-Source NVK Driver

    Phoronix: NVIDIA Developer Opens Feature Pull Request For Open-Source NVK Driver

    If your interest didn't pique enough when the former Nouveau lead developer joined NVIDIA and sent out a big patch series for this originally-reverse-engineered, open-source NVIDIA kernel driver, here's another plot twist: another NVIDIA engineer opening a merge request adding to the Mesa NVK Vulkan driver...

    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
    Maybe nvk is valve requirement for a new.generation of a gaming console (not portable)?

    Comment


    • #3
      2024 is the year of the NVIDIA Linux desktop.

      Comment


      • #4
        Originally posted by pharmasolin View Post
        Maybe nvk is valve requirement for a new.generation of a gaming console (not portable)?
        The only console using Nvidia is Switch. May be Nintendo is pressuring them?

        Comment


        • #5
          Originally posted by pharmasolin View Post
          Maybe nvk is valve requirement for a new.generation of a gaming console (not portable)?
          i am pretty sure that if valve build a non-portable gaming console then they will use
          Strix Point (STX) and Strix Halo (STX Halo).

          TSMC N4 TSMC N4P
          Multi-Chiplet-Module Monolithisch
          Zen 5 + Zen 5c Zen 5 + Zen 5c
          bis zu 16C/32T bis zu 12C/24T
          16x Zen5
          oder
          8x Zen 5 + 8x Zen 5c
          4x Zen + 8x Zen 5c
          RDNA 3.5 RDNA 3.5
          bis zu 40 CUs bis zu 16 CUs
          bis zu 2.560 Shader bis zu 1.024 Shader
          64 MiByte 24 - 32 MiByte
          256-Bit LPDDR5X
          und
          DDR5
          128-Bit LPDDR5X
          und
          DDR5
          XDNA 2 XDNA 2
          45 - 50 TOPS 45 - 50 TOPS
          55 - 120 Watt 28 - 45 Watt
          und
          15 - 30 Watt
          ​
          Phantom circuit Sequence Reducer Dyslexia

          Comment


          • #6
            Yet more proof Nvidia doesn't hate OSS. They just have a business to run and they embrace it when it makes sense for them to do so.

            Comment


            • #7
              The reason is simple:

              1) Ignoring Linux users like they're just activist morons: No dev cost, Not a lot of GPUs sold to them.
              2) Showing the bare minimum, paying ONE engineer to contribute time to time: A few more GPU to be sold.

              CFO choice : Extra profit possible with minimum extra cost ? B is fine.

              Comment


              • #8
                Originally posted by bug77 View Post
                Yet more proof Nvidia doesn't hate OSS. They just have a business to run and they embrace it when it makes sense for them to do so.
                This is certainly an improvement but I'm still highly pessimistic, especially with the CUDA license change. My prediction is that they will reluctantly allow FOSS 3D (Vulkan/OpenGL) while keeping compute/CUDA very much proprietary. That is what will generate the most income. If they get the chance to close everything back up, they'll take it in a heartbeat. Their developers seem like good people, the problem is the higher ups.

                Comment


                • #9
                  Originally posted by bug77 View Post
                  Yet more proof Nvidia doesn't hate OSS. They just have a business to run and they embrace it when it makes sense for them to do so.
                  I think they are simply practical.

                  Old Nouveau sucked as base for potential open source driver so they didn't see future there and they know they cannot just opensource propertiary driver. But current NVK+Nova+zink means essentially, they get some kind of kernel driver, they get modern Vulkan driver, and can forget largely about legacy openGL support.

                  The only question is how far they will go with CUDA. Will there be some propertiary user mode CUDA driver hooked to opensource normal driver or will there be something truly open source as CUDA driver or simply no CUDA at all for open source stack. There could be also something crazy like CUDA to Vulkan translation layer with bunch of special Vulkan extensions.

                  Comment


                  • #10
                    Originally posted by rmfx View Post
                    The reason is simple:

                    1) Ignoring Linux users like they're just activist morons: No dev cost, Not a lot of GPUs sold to them.
                    2) Showing the bare minimum, paying ONE engineer to contribute time to time: A few more GPU to be sold.

                    CFO choice : Extra profit possible with minimum extra cost ? B is fine.
                    Highly unlikely for them to focus on small numbers. You are talking about a trillion dollars company, so they are unlikely to focus on small extra profit from extra desktop GPU sales.

                    My guess is that they fear bigger consequences of ignoring Linux. The cloud runs Linux, and AI is all happening in the cloud. If cloud maintainers get fed up with out-of-tree drivers they will have more incentive to look for alternatives.

                    Comment

                    Working...
                    X