Announcement

Collapse
No announcement yet.

AMD Linux Graphics: The Latest Open-Source RadeonSI Driver Moves On To Smacking Catalyst

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

  • #21
    I'm honestly shocked at the Unigine Valley bench, in a very good way.

    Comment


    • #22
      Originally posted by marek View Post
      I wonder why the Catalyst rendering allegedly looks better in Valley. Can somebody paste the Valley log on Catalyst to see what OpenGL extensions it's using?
      Here is the data on my system (Valley 64-bit with Extreme preset @1680*1050). That's on r600g, so maybe not what you are looking for though. I bolded the 3 extensions advertised with fglrx and not mesa.

      Fglrx 15.7 on CentOS 6.6
      ******************************
      ---- Render ----
      Renderer: ATI R800 1024MB
      OpenGL vendor: ATI Technologies Inc.
      OpenGL renderer: AMD Radeon HD 5800 Series
      OpenGL version: 3.2.13397 Core Profile Context 15.20.1046
      OpenGL flags: Core Profile
      Found required GL_ARB_map_buffer_range
      Found required GL_ARB_vertex_array_object
      Found required GL_ARB_draw_instanced
      Found required GL_ARB_draw_elements_base_vertex
      Found required GL_ARB_transform_feedback
      Found required GL_ARB_half_float_vertex
      Found required GL_ARB_half_float_pixel
      Found required GL_ARB_framebuffer_object
      Found required GL_ARB_texture_multisample
      Found required GL_ARB_uniform_buffer_object
      Found required GL_ARB_geometry_shader4
      Found optional GL_ARB_blend_func_extended
      Found optional GL_ARB_tessellation_shader
      Found optional GL_ARB_shader_bit_encoding
      Found optional GL_ARB_sample_shading
      Found optional GL_ARB_compute_shader
      Found optional GL_ARB_gpu_shader5
      Found optional GL_EXT_texture_compression_s3tc
      Found optional GL_ARB_texture_compression_rgtc
      Shading language: 4.40
      Maximum texture size: 16384
      Maximum texture units: 108
      Maximum texture renders: 8




      Mesa 11.1.0-devel (git-d38a560) on Fedora 22
      ******************************
      ---- Render ----
      GLRender::GLRender(): Unknown GPU
      OpenGL vendor: X.Org
      OpenGL renderer: Gallium 0.4 on AMD CYPRESS (DRM 2.42.0)
      OpenGL version: 3.3 (Core Profile) Mesa 11.1.0-devel (git-d38a560)
      OpenGL flags: Core Profile
      Found required GL_ARB_map_buffer_range
      Found required GL_ARB_vertex_array_object
      Found required GL_ARB_draw_instanced
      Found required GL_ARB_draw_elements_base_vertex
      Found required GL_ARB_transform_feedback
      Found required GL_ARB_half_float_vertex
      Found required GL_ARB_half_float_pixel
      Found required GL_ARB_framebuffer_object
      Found required GL_ARB_texture_multisample
      Found required GL_ARB_uniform_buffer_object
      Found required GL_ARB_geometry_shader4
      Found optional GL_ARB_blend_func_extended
      Found optional GL_ARB_shader_bit_encoding
      Found optional GL_ARB_sample_shading
      Found optional GL_EXT_texture_compression_s3tc
      Found optional GL_ARB_texture_compression_rgtc
      Shading language: 3.30
      Maximum texture size: 16384
      Maximum texture units: 48
      Maximum texture renders: 8

      Comment


      • #23
        sweet jesus and mary jane!!!! these are some sweet improvements

        if i'm correct with assumption that metro uses newest GL from all other tests, it might just be time to profile it and see where slowdown happened compared to catalyst. based on anihilation in other tests, it could as well point out very few, but very meaningful performance needed improvements

        Comment


        • #24
          whitecat: I'm sure Valley doesn't use tessellation, but I don't know if it uses a compute shader. The log only tells us what the engine "can" use, not what the engine is using.

          Another thing that may explain the difference is drirc. This is how /etc/drirc should look like:


          If it's different, we know where the problem is. The file was changed with Mesa 11.0.

          Comment


          • #25
            Originally posted by marek View Post
            I wonder why the Catalyst rendering allegedly looks better in Valley. Can somebody paste the Valley log on Catalyst to see what OpenGL extensions it's using?

            The open source driver uses the best quality rendering it can. No compromises.
            This is Render subsection log from Valley:
            ---- Render ----
            GLRender::GLRender(): Unknown ATI GPU
            OpenGL vendor: ATI Technologies Inc.
            OpenGL renderer: AMD Radeon(TM) R7 Graphics
            OpenGL version: 3.2.13374 Core Profile Context 15.20.1046
            OpenGL flags: Core Profile
            Found required GL_ARB_map_buffer_range
            Found required GL_ARB_vertex_array_object
            Found required GL_ARB_draw_instanced
            Found required GL_ARB_draw_elements_base_vertex
            Found required GL_ARB_transform_feedback
            Found required GL_ARB_half_float_vertex
            Found required GL_ARB_half_float_pixel
            Found required GL_ARB_framebuffer_object
            Found required GL_ARB_texture_multisample
            Found required GL_ARB_uniform_buffer_object
            Found required GL_ARB_geometry_shader4
            Found optional GL_ARB_blend_func_extended
            Found optional GL_ARB_tessellation_shader
            Found optional GL_ARB_shader_bit_encoding
            Found optional GL_ARB_sample_shading
            Found optional GL_ARB_compute_shader
            Found optional GL_ARB_gpu_shader5
            Found optional GL_EXT_texture_compression_s3tc
            Found optional GL_ARB_texture_compression_rgtc
            Shading language: 4.40
            Maximum texture size: 16384
            Maximum texture units: 192
            Maximum texture renders: 8
            I used A10-7850K + 8GB 2166MHz RAM + Catalyst 15.7 for this benchmark. I set 1920x1080 resolution + High quality settings (AA off). My result is following:
            FPS: 12.2
            Score: 511
            Min FPS: 7.1
            Max FPS: 21.2

            Comment


            • #26
              Originally posted by marek View Post
              Another thing that may explain the difference is drirc. This is how /etc/drirc should look like:
              http://cgit.freedesktop.org/mesa/mes...i/common/drirc
              I never enabled any drirc config when running Unigine benchmarks. Should I ? I mean, is that mandatory in order to expect having a "normal" behavior with these benchmarks ?

              Comment


              • #27
                IIRC the catalyst driver had some kind of limit in some cases (see catalyst vs nVidia comparisons). maybe RadeonSI will now remove it.
                on another note: there will come the benefits from kernel scheduler and graphics scheduler on top??? that would be incredibly neat!

                Comment


                • #28
                  so that's decided
                  my next card will be a radeon

                  Comment


                  • #29
                    Originally posted by whitecat View Post
                    I never enabled any drirc config when running Unigine benchmarks. Should I ? I mean, is that mandatory in order to expect having a "normal" behavior with these benchmarks ?
                    The benchmarks just don't work without drirc.

                    Comment


                    • #30
                      I guess next time will be that there was rendering regression in mesa article

                      Those R9 290 results in Valley looks like in line with GTX 980... Can you Michael write article how opensource radeon driver smashed nvidia driver

                      Originally I was going to use the latest Catalyst 15.7 driver release, but its kernel module was running into issues on this Intel Skylake-based system used for testing on Ubuntu 15.04. Thus I resorted to using the packaged fglrx 15.20.2 / OpenGL 4.4.13374 OpenGL driver as packaged in Ubuntu Vivid that would play fine with the Skylake system running Ubuntu with the Xfce desktop.
                      Or maybe this is all about that some things behave weird on Skylake?
                      Last edited by dungeon; 01 September 2015, 05:47 PM.

                      Comment

                      Working...
                      X