Announcement

Collapse
No announcement yet.

A Fresh Look At The AMD Radeon Gallium3D Performance

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

  • #31
    Yeah, life is funny sometimes.

    Comment


    • #32
      Originally posted by deanjo View Post
      *snicker* and yet it is so easily defeated.


      The funniest bit is that even if it was somehow magically perfected, people would get around it just fine by using a large HDTV and record using an expensive camcorder.

      Sure, it wouldn't match the original quality, but it would probably be "good enough" for the vast majority of what's pirated. Just get on any torrent site and take a look at how many TS (hidden shaky camcorder in a movie theater) videos there are.
      Last edited by smitty3268; 06-10-2011, 11:46 PM.

      Comment


      • #33
        Finally I can see some useful 3D benchmark numbers with performance features enabled. Thank you, Michael.

        I think a really relevant benchmark would be Minecraft (www.minecraft.net), since it is quite a popular game right now (will leave public beta on 11-11-2011) and has a native linux client. It works even on my centrino/rv350 system with ubuntu natty r300g, but only in a small window and with reduced distance and fog settings.
        I'd be very interested to see how it runs on AMDs 4250, 6570, 5770, and 5850 with r600g + performance features.

        Comment


        • #34
          Originally posted by smitty3268 View Post
          The funniest bit is that even if it was somehow magically perfected, people would get around it just fine by using a large HDTV and record using an expensive camcorder.
          If the MPAA had its way, that camcorder would be required by law to stop recording and delete the file when it saw a watermark. I'm not kidding; this sort of system was seriously proposed.

          Comment


          • #35
            Originally posted by bridgman View Post
            Yeah, life is funny sometimes.
            Bridgman, what is expected speed boost when HiZ gets implemented?

            Comment


            • #36
              Apples to Apples

              Isn't disabling swapbufferswait and enabling color-tiling a cheat?

              IMO only defaults matter and I want to know if the Catalyst driver has these options by default.

              Comment


              • #37
                Originally posted by zoomblab View Post
                Isn't disabling swapbufferswait and enabling color-tiling a cheat?

                IMO only defaults matter and I want to know if the Catalyst driver has these options by default.
                What sort of a question is that?

                Of course Catalyst does colour tiling, in fact, it does much more complex colour tiling and it does texture tiling too.

                And in any benchmark where the FPS is higher than the refresh rate (usually 60 fps), there is obviously no swap buffer waiting.

                Previous benchmarks were unfair because important optimizations were not used by the open drivers.
                Last edited by pingufunkybeat; 06-11-2011, 07:21 AM.

                Comment


                • #38
                  Originally posted by Flynn View Post
                  and has a native linux client.
                  Yes and No its Java

                  Comment


                  • #39
                    These are very nice results, BTW.

                    AFAIK, we are still missing 2d tiling, texture tiling and Hyper-Z. With these goodies, we should be more competitive.

                    Comment


                    • #40
                      Blender 2.5+ benchmarks please

                      Originally posted by EvilTwin View Post
                      how about benching the blender view-port performance in different modes?
                      this is something you actually use compared to benches like lightsmark
                      By viewport do you mean Blender Game Engine or just.. viewport?

                      Well, anyway I would be very interested in seeing some Gallium vs. blobs benchmarks that measure BGE 2.5+. In fact I might be willing to write some test profiles myself eventually (if I find the time).

                      Comment


                      • #41
                        Originally posted by pingufunkybeat View Post
                        And in any benchmark where the FPS is higher than the refresh rate (usually 60 fps), there is obviously no swap buffer waiting.
                        I doubt it. That would mean tearing with catalyst by default, which is unlikely. I admit, however, I have no ATI hardware to see for myself.

                        Comment


                        • #42
                          Originally posted by ahlaht View Post
                          By viewport do you mean Blender Game Engine or just.. viewport?

                          Well, anyway I would be very interested in seeing some Gallium vs. blobs benchmarks that measure BGE 2.5+. In fact I might be willing to write some test profiles myself eventually (if I find the time).
                          just the view-port (solid mode, edit mode, sculpt mode and so on)
                          i personally don't use the game engine
                          but even then you need to create content first

                          Comment


                          • #43
                            Originally posted by Kano View Post
                            the oss driver support is very basic when you compare opengl features. xvba nobody really needs without full h264 l5.1 support. if you want to play a game (or at least look at the heaven demo) with tesselation then you are lost without binary driver. it is fun to test oss drivers, with my intel atom netbook with gma 945 even gl2benchmark runs correctly. fbo (test 4) does not work with snb and other oss drivers do not even show the reference pictures correctly. i really would like to see it running with every oss driver.

                            http://kanotix.com/files/hellfire/contrib/gl2benchmark/
                            I did not find any bug reports about gl2benchmark. Well, with the only exception being the (now resolved) bug I filed the last time you whined about it...

                            Comment


                            • #44
                              The other issue with opening up the closed source driver is that the coding styles are VERY different and there are abstraction layers in various modules so that we can share the code with other OSes. This won't fly in the kernel. The driver would have to go into staging and it would be a huge amount of work to change it to match the kernel style guidelines. Plus, once the coding style got changed, it would be nearly impossible to share code between OSes since the two code bases would be so different. So that defeats the main purpose of sharing code.

                              Comment


                              • #45
                                Originally posted by zoomblab View Post
                                I doubt it. That would mean tearing with catalyst by default, which is unlikely. I admit, however, I have no ATI hardware to see for myself.
                                It's true. Vsync is set to "application controlled, default off" in the Ati and Nvidia blobs, whereas it's enabled in the open-source drivers.

                                Comment

                                Working...
                                X