Announcement

Collapse
No announcement yet.

Open-Source Radeon Vulkan Driver Improvement Scores Huge Ray-Tracing Wins

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

  • Open-Source Radeon Vulkan Driver Improvement Scores Huge Ray-Tracing Wins

    Phoronix: Open-Source Radeon Vulkan Driver Improvement Scores Huge Ray-Tracing Wins

    A change merged today for Mesa 24.0 is yielding much better Vulkan ray-tracing performance for the Radeon Vulkan driver "RADV" across a number of games...

    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
    Great work!

    It would be really interesting to compare this against Windows

    Comment


    • #3
      Wow that is a huge improvement in RT rendering. This might get me to finally upgrade my video card to something modern.

      Comment


      • #4
        How about image quality? Does this cut any corner for that gain in performance?

        Comment


        • #5
          Originally posted by M@GOid View Post
          How about image quality? Does this cut any corner for that gain in performance?
          I would not be surprised if it lowered the precision of the calculations a bit.

          Probably nothing anyone would notice during actual game play.

          Comment


          • #6
            I wonder if it will improve the Deck's RT performance as well.
            ​​​​​For games like Doom Eternal.

            Comment


            • #7
              We need to keep these claims in perspective, there is a claimed increase of 140% more frames per second in Ghostwire: Tokyo but the actual increase is from 5 FPS to 12 FPS.

              It reminds me of The Office when Pam gets an award of something like that and the other sales people are complaining, her boyfriend is is now co-manager says that he gave it to her because she doubled her sales and one of the sales people asks sarcastically "What, from 2 to 4?" and it cuts to Pam looking into the camera and saying "Yup!".

              Same thing with Sackboy they brag about a 75% increase but the actual numbers are from 8 FPS to 14 FPS.

              The games are still not playable.

              Comment


              • #8
                Originally posted by Kjell View Post
                Great work!

                It would be really interesting to compare this against Windows
                I get on my 7900XT:
                Witcher 3 - Linux 60fps Windows 66fps
                Dying Light 2 - Linux 78fps Windows 96fps
                Desordre (Path Tracing Low) - Linux 75fps WIndows 75fps
                Quake II RTX - Linux 72fps Windows 74fps

                I'm not sure about that Quake result. If I use the AMDVLK drivers I get 82fps for Quake II RTX on Linux, so even faster than Windows. The others are identical setups though. It would be interesting to see this tested properly.

                Comment


                • #9
                  Originally posted by sophisticles View Post
                  We need to keep these claims in perspective, there is a claimed increase of 140% more frames per second in Ghostwire: Tokyo but the actual increase is from 5 FPS to 12 FPS.

                  It reminds me of The Office when Pam gets an award of something like that and the other sales people are complaining, her boyfriend is is now co-manager says that he gave it to her because she doubled her sales and one of the sales people asks sarcastically "What, from 2 to 4?" and it cuts to Pam looking into the camera and saying "Yup!".

                  Same thing with Sackboy they brag about a 75% increase but the actual numbers are from 8 FPS to 14 FPS.

                  The games are still not playable.
                  Geez. Perspective is one way to put it. It could be that those cards just plain suck with Ghostwire and Sackboy because those games require more beefier cards or something in their code is garbage. Don't crap all over Friedrich's contribution even if in a couple cases it was a "meh" improvement. Free performance is free performance. Though I won't be shocked if someone looks at his code and figures out a way to squeeze even more out of it because he may not have thought of something that some other code ninja will. We've seen that time and again.

                  Comment


                  • #10
                    Originally posted by M@GOid View Post
                    How about image quality? Does this cut any corner for that gain in performance?
                    No. The rendering results before/after this MR are exactly the same. If there are any changes that's a bug, but I haven't noticed any in my testing.

                    Comment

                    Working...
                    X