Announcement

Collapse
No announcement yet.

ARB_copy_image Support Lands For AMD RadeonSI

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

  • #11
    I played Bioshock: Infinity with radeonsi and the performance was very bad. Turning video options on and off makes little difference. Maybe this patch helps with something.

    Comment


    • #12
      Maybe you can get +15% with this like on Catalyst (just a guess), but game is also CPU capped - so crying for threaded GL too, well that is something mesa drivers does not have currently.

      But it should be comparable maybe 80% of Catalyst without mentioned game switch and without profile which enable threaded GL.
      Last edited by dungeon; 10-28-2015, 11:59 AM.

      Comment


      • #13
        Originally posted by [email protected] View Post
        I played Bioshock: Infinity with radeonsi and the performance was very bad. Turning video options on and off makes little difference. Maybe this patch helps with something.
        did you add this fix:
        https://www.reddit.com/r/linux_gamin...g_and_texture/

        I wonder if michael ever used that fix for his benchmarks. It works for all unreal engine 3 games afaik

        Comment


        • #14
          It seems like the implementation can be more optimized.

          Originally posted by Azpegath View Post
          Hurray! Finally RadeonSI is in the lead over Nvidia
          You mean over three years after Nvidia. This feature was initially designed by Nvidia.

          Comment


          • #15
            I propose to drop benchmarking of OpenGL and also stop playing OpenGL games. D3D and OGL wile they target the same Shader Models, they do things differently. D3D is the unrestricted one (on threading and other staff), it is unstable for that reason and gives more responsibility to game devs (it's for gaming). OGL is the restricted one (many things can be done only together), it is stable with less performance (only for CAD), letter unrestricted extensions they break OGL. This happened because corporations wanted gaming under specific platforms only, wile the two graphics vendors can keep pro graphics for them. They where also many game devs that they wanted to use OGL before WinXP become the top OS and the two gpu vendors let them down.

            Catalyst usually has 2-3 times the overhead of Nvidia and Gallium (RadeonSI) for that reason. Nvidia did a small difference from the PS3 times when they had to use OGL for actual gaming, MESA follows. When i run PCSX2 and FFX-2usa on my Xubuntu Laptop with i3-3110m and Radeon8690, i get 20fps with Catalyst and a throttled CPU and 60fps with RadeonSI. When i run FFXIII with Wine, i get 5fps with Catalyst, 8fps with Catalyst CSMT, 14fps with RadeonSI regardless of CSMT and 17 FPS with NINE, the game is single thread so i don't now if NINE with Extra Threaded Optimizations can give the 20-25fps that i get with Windowz or more.

            Anyway just benchmark the Windows binary with Nine for now (add also that Wine has 20% overhead regardless), and then Native titles with Vulcan. OGL isn't useful for more than a dozen games and benchmarking it doesn't show anything.
            Last edited by artivision; 10-28-2015, 02:11 PM.

            Comment


            • #16
              Originally posted by artivision View Post
              I propose to drop benchmarking of OpenGL and also stop playing OpenGL games.
              ...
              OGL is the restricted one (many things can be done only together), it is stable with less performance (only for CAD), letter unrestricted extensions they break OGL.
              ...
              OGL isn't useful for more than a dozen games and benchmarking it doesn't show anything.
              Clearly not an OpenGL dev, as your claims are wrong. Nvidia manages to provide excellent OpenGL stability and performance across platforms, and even outperform Direct3D. The fact that all other drivers are partly or largely unacceptable is due to lack of effort. Not to mention that Gallium drivers adds an extra layer of abstraction on top.

              Comment


              • #17
                Originally posted by efikkan View Post
                It seems like the implementation can be more optimized.
                You mean over three years after Nvidia. This feature was initially designed by Nvidia.
                I mean "Nvidia" as in noveau drivers. See MesaMatrix.

                Comment


                • #18
                  Originally posted by efikkan View Post
                  Clearly not an OpenGL dev, as your claims are wrong. Nvidia manages to provide excellent OpenGL stability and performance across platforms, and even outperform Direct3D. The fact that all other drivers are partly or largely unacceptable is due to lack of effort. Not to mention that Gallium drivers adds an extra layer of abstraction on top.
                  Sorry but you are wrong. Nvidia OGL gives only 70% of the D3D frame rate (see Unigine on Windowz). GL Threaded Optimizations usually break OGL. Gallium doesn't add extra layers, all drivers work the same way: On Catalyst D3D or OGL state_trackers convert data to a low level Api (AMD_API) and to an assembly called AMD_IL, then to HW just like Gallium. The same is Nvidia.

                  Comment


                  • #19
                    Originally posted by artivision View Post
                    I propose to drop benchmarking of OpenGL and also stop playing OpenGL games.
                    I have Playstation 4 for that purpose - with that i am free of all PC bullshit APIs sometimes

                    Pros: There is no better API then API maded for any single GPU
                    Cons: It is tied to one single GPU
                    Last edited by dungeon; 10-28-2015, 02:48 PM.

                    Comment


                    • #20
                      Originally posted by artivision View Post

                      Sorry but you are wrong. Nvidia OGL gives only 70% of the D3D frame rate (see Unigine on Windowz). GL Threaded Optimizations usually break OGL. Gallium doesn't add extra layers, all drivers work the same way: On Catalyst D3D or OGL state_trackers convert data to a low level Api (AMD_API) and to an assembly called AMD_IL, then to HW just like Gallium. The same is Nvidia.
                      OpenGL is not the constrained API, though. It's the low-level one that lets you do whatever you want. D3D is the constrained API that forces you to do things in particular ways. That makes it easier for drivers to optimize, since they better know how applications are going to use the driver.

                      Comment

                      Working...
                      X