Announcement

Collapse
No announcement yet.

Radeon Gallium3D: A Half-Decade Behind Catalyst?

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

  • #31
    Originally posted by bridgman View Post
    IIRC the thinking was "the code is in a CM system, so you can go back and create a branch from any point you want if needed", ie the branch doesn't have to be created today. All previous versions of the code are in the CM system and available by rolling back to an earlier commit.
    Really, all I want is not to have what happened to Catalyst (to the R300 and earlier stuff) to happen to the old Mesa drivers. IDK, silly really.

    Comment


    • #32
      Originally posted by LinuxID10T View Post
      Unless you are running a distro that doesn't support it or are running Solaris/BSD. I mean, why not just leave the code in a branch as unmaintained. I mean, it would just be nice to have if you are running said older hardware.
      But they won't build and no one is updating them. Why keep it in the tree? They are readily accessible locally via git or old tarballs or on the web via cgit. Keeping them around, even if they are not being built just causes problems when things are changed or new features are added since you end up having to wade through extra confusing deprecated crap.

      Comment


      • #33
        Is there a plan to remove radeon and r200 from mesa? Today mesa-git needs explicit shared dricore - dricore never worked for me.

        Athlon XP 2200+ 1.8GHz Radeon 9250 128MB 128-bit, mesa git KMS with kernel 3.1.10 vs mesa 7.5.2 UMS kernel 2.6.32.54.

        Code:
        Open Arena 0.8.5
         
                  640x480 800x600 1024x600 1024x768 1280x720 1280x768 1280x800 1366x768 1280x960 1440x900 1280x1024 1600x900 1400x1050 1680x1050 1920x1080
        
        UMS         77.3   76.3     76.1     72.4     66.8     64.0     62.2     61.0     54.5     51.9      47.6      52.0     48.4     40.7      35.2
        
        KMS         51.8   51.9     52.1     51.3     50.2     49.2     48.3     47.8     44.0     42.2      42.2      39.2     38.3     33.3      29.5
        
        
        Urban Terror 4.1
        
                  640x480 800x600 1024x600 1024x768 1280x720 1280x768 1280x800 1366x768 1280x960 1440x900 1280x1024 1600x900 1400x1050 1680x1050 1920x1080
        
        UMS        141.7   115.2    96.9     80.0     70.6     67.0     64.8     63.3     55.5     54.7      52.6      49.8     48.9     41.3      35.4   
        
        KMS        107.8   100.2    89.6     74.0     66.4                                                                      45.2               33.3

        Comment


        • #34
          With the latest mesa you can enable tiling for KMS as well on r1xx/r2xx. Option "ColorTiling" "true" in the device section of your xorg.conf. UMS already has tiling enabled on older asics.

          Comment


          • #35
            Originally posted by dungeon View Post
            Is there a plan to remove radeon and r200 from mesa?
            No. There are no plans to remove radeon and r200 from mesa.

            Comment


            • #36
              Yes i know i have enabled ColorTilling. Do you know why shared dricore not work for r200?

              P.S. Results are with CT enabled - this card and driver is better than fglrx in this tests.
              Last edited by dungeon; 01-21-2012, 01:43 PM.

              Comment


              • #37
                As may I concern catalyst 9.3 is perfectly compatible with ubuntu 9.04 (you can make the packages for that distro)

                I have an X1650 pro AGP that been working with 8.04.4 by using 9.3 catalyst. It worked well and regnum online was playable at 40 fps. (mesa 8 in kubuntu 11.10 have 16 fps)

                In a few days, I will try heron again with catalyst 8.3 (I remember that jockey gave me blank screen after using repository drivers)

                nice review!

                Comment


                • #38
                  Originally posted by dungeon View Post
                  Is there a plan to remove radeon and r200 from mesa? Today mesa-git needs explicit shared dricore - dricore never worked for me.

                  Athlon XP 2200+ 1.8GHz Radeon 9250 128MB 128-bit, mesa git KMS with kernel 3.1.10 vs mesa 7.5.2 UMS kernel 2.6.32.54.

                  Code:
                  Open Arena 0.8.5
                   
                            640x480 800x600 1024x600 1024x768 1280x720 1280x768 1280x800 1366x768 1280x960 1440x900 1280x1024 1600x900 1400x1050 1680x1050 1920x1080
                  
                  UMS         77.3   76.3     76.1     72.4     66.8     64.0     62.2     61.0     54.5     51.9      47.6      52.0     48.4     40.7      35.2
                  
                  KMS         51.8   51.9     52.1     51.3     50.2     49.2     48.3     47.8     44.0     42.2      42.2      39.2     38.3     33.3      29.5
                  
                  
                  Urban Terror 4.1
                  
                            640x480 800x600 1024x600 1024x768 1280x720 1280x768 1280x800 1366x768 1280x960 1440x900 1280x1024 1600x900 1400x1050 1680x1050 1920x1080
                  
                  UMS        141.7   115.2    96.9     80.0     70.6     67.0     64.8     63.3     55.5     54.7      52.6      49.8     48.9     41.3      35.4   
                  
                  KMS        107.8   100.2    89.6     74.0     66.4                                                                      45.2               33.3
                  I was not using XP for 5 years now but intalled it to test this old graphic card... and OMG i even forgot how fast it can be.

                  Code:
                  Open Arena 0.8.5
                  
                  640x480 800x600 1024x768 1280x720 1280x768 1280x960 1280x1024 1920x1080
                  
                   128.2   127.5   121.0    113.1    108.3     91.7      87.1     57.0
                  
                  Urban Terror 4.1
                  
                  640x480 800x600 1024x768 1280x720 1280x768 1280x960 1280x1024 1920x1080
                  
                   200.3   188.7   142.0    124.2    117.8     98.2      92.9     61.0

                  Comment


                  • #39
                    I tested UT also on Debian Etch, so squash results...
                    Code:
                    Urban Terror 4.1
                    
                                640x480 800x600 1024x768 1280x1024 1920x1080
                    
                    UMS-7.5.2    141.7   115.2    80.0     52.6      35.4   
                    
                    KMS-git      107.8   100.2    74.0               33.3
                    
                    FGLRX        216.1   171.2    114.7    72.6      48.5
                    
                    UMS-6.5.1    146.2   118.5    82.5     53.3
                    
                    WindowsXP    200.3   188.7   142.0     92.9      61.0
                    DVI was not work with fglrx, but radeon was with both VGA/DVI until 3D is on with DVI then monitor go to sleep - just like with current driver.

                    Fglrx on Windows and Catalyst on Windows both give blurred fonts on VGA, DVI on Windows is OKish - on Linux both are good (with bug on DVI).

                    Video playback is good with VideoOverlay on, direct3d on Windows, Overlay on UMS and KMS's only TexturedVideo explicitly needs compozite naci-vblank (dia-tear) which halfed vsynced 3d;(.
                    ..............

                    Need to change cables (etch/fglrx was not work when both are pluged in), i'll become blind with VGA on XP.
                    Last edited by dungeon; 01-24-2012, 08:56 PM.

                    Comment


                    • #40
                      Hi, I have just tried a simple glxgears test with all the desktop environments I have. I have to say that the results are quite astonishing. It proves that the testing methodology of the article is flawed and the presented results are unfair to say the least. Now the comparison (rounded to tens):

                      Unity - 4260 FPS
                      Unity 2D - 3540 FPS
                      KDE - 3830 FPS
                      Gnome 2 - 6000 FPS
                      Xfce - 6160 FPS

                      The article is comparing Catalyst and R300g and at the same time it's comparing Unity and Gnome 2 (why?). We know from the article that Catalyst+Gnome2 is faster than R300g+Unity. We can say for sure that Catalyst is likely faster than R300g in most or all tests, but we already know that from previous benchmarks. We also know that Gnome 2 is faster than Unity. What the article doesn't show is how much exactly Catalyst is faster than R300g, because it doesn't compare the two in the same desktop environment.

                      Michael, next time you make an article, please do it at least right.

                      Comment


                      • #41
                        Why not run the benchmarks in a separate xserver instance? There are tutorials/scripts all over the net showing how to do it.

                        Comment


                        • #42
                          Marek can you please test kde without desktop effects? Thanks
                          ## VGA ##
                          AMD: X1950XTX, HD3870, HD5870
                          Intel: GMA45, HD3000 (Core i5 2500K)

                          Comment


                          • #43
                            Originally posted by marek View Post
                            Hi, I have just tried a simple glxgears test with all the desktop environments I have. I have to say that the results are quite astonishing. It proves that the testing methodology of the article is flawed and the presented results are unfair to say the least. Now the comparison (rounded to tens):

                            Unity - 4260 FPS
                            Unity 2D - 3540 FPS
                            KDE - 3830 FPS
                            Gnome 2 - 6000 FPS
                            Xfce - 6160 FPS

                            The article is comparing Catalyst and R300g and at the same time it's comparing Unity and Gnome 2 (why?). We know from the article that Catalyst+Gnome2 is faster than R300g+Unity. We can say for sure that Catalyst is likely faster than R300g in most or all tests, but we already know that from previous benchmarks. We also know that Gnome 2 is faster than Unity. What the article doesn't show is how much exactly Catalyst is faster than R300g, because it doesn't compare the two in the same desktop environment.

                            Michael, next time you make an article, please do it at least right.
                            glxgears should not used as a benchmark but yes you are right michael compare different stuff.

                            he should test only x+game without gnome or kde.

                            KDE also improve a lot if you don't use the catalyst with kde4,8 you get a much faster openGL2 rendering. with the newest QT its also much faster.

                            Comment


                            • #44
                              Originally posted by marek View Post
                              Michael, next time you make an article, please do it at least right.
                              Unfortunately, better articles are not economically viable at this point.

                              Comment


                              • #45
                                Originally posted by AnonymousCoward View Post
                                Unfortunately, better articles are not economically viable at this point.
                                why? i can not imagine any criteria.

                                Comment

                                Working...
                                X