Announcement

Collapse
No announcement yet.

ATI R600/700 OSS 3D Driver Reaches Gears Milestone

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

  • Originally posted by pingufunkybeat View Post
    I didn't forget it -- especially since the support for my chipset was discontinued so my laptop didn't profit from all the patches nvidia made to fix the problems.
    Oh... that's got to suck.

    I stopped using nVidia when they pulled that trick on Compiz users, giving everyone the black window bug (when previously it just worked) then announcing they'll drop Geforce MX support the next day.

    Comment


    • Originally posted by energyman View Post
      nvidia fucked up so hard I switched over to AMD.
      And look where that got you.

      Comment


      • Well, you have to admit that nVidia make a mean blob. The driver is not bad.

        But that episode proved to me very clearly that things are not perfect in blob-land, and that open-source drivers would have been preferable.

        So I bought an ATi card for my new desktop. The 3D is not ready yet, and it will be a while before it can compare to nVidia's blob, but there is something soothing for a soul when you run free drivers on open hardware.

        Comment


        • Originally posted by pingufunkybeat View Post
          Well, you have to admit that nVidia make a mean blob. The driver is not bad.

          But that episode proved to me very clearly that things are not perfect in blob-land, and that open-source drivers would have been preferable.

          So I bought an ATi card for my new desktop. The 3D is not ready yet, and it will be a while before it can compare to nVidia's blob, but there is something soothing for a soul when you run free drivers on open hardware.
          exactly

          Originally posted by RealNC View Post
          And look where that got you.
          see pingufunkybeat's posting. Also: fglrx works well enough for my needs.

          Comment


          • One of the recent commits seems to have improved the graphical corruption issue.

            OpenArena is still unplayable, but much better than before, and supertux in OpenGL mode and Neverball are both much better than before (although still showing slight corruption).

            Great work.

            glxgears gives similar fps as before (slightly lower), but I get this message:

            Code:
            *********************************WARN_ONCE*********************************
            File radeon_dma.c function radeonReleaseDmaRegions line 302
            Leaking dma buffer object!
            ***************************************************************************
            EDIT: KWin still hardlocks the computer when it tries to use OpenGL compositing. Foobillard too.
            Last edited by pingufunkybeat; 18 August 2009, 07:43 PM.

            Comment


            • AFAIK current thinking is that this commit was the one that helped with corruption :



              If I understand the change correctly it keeps used buffers around for a while rather than re-using them immediately. That fits with the observation that corruption is much reduced but not eliminated... which, in turn, implies that the root problem may be that we are re-using buffers too soon, before the GPU has finished reading from them.
              Last edited by bridgman; 18 August 2009, 08:13 PM.
              Test signature

              Comment


              • It seems to me that more complex games suffer more corruption than the simple games. Stuff like supertux is almost perfect, with a flashing triangle here or there, Neverball has more flickering, but is completely playable, while OpenArena looks like a depressed Picasso did the level design

                Anyway, I believe that you devs know how to reproduce the problem and are looking for the fix. Keep up the good work, and let us know if we can help with testing or otherwise.

                If I understand the change correctly it keeps used buffers around for a while rather than re-using them immediately. That fits with the observaction that corruption is much reduced but not eliminated.
                If I'm not mistaken, there was far less corruption back when memory copies were expensive and everything slow as a result. The 100x speedup brought by the accelerated buffer copy seems to have brought the corruption, at least according to the posters here.
                Last edited by pingufunkybeat; 18 August 2009, 08:19 PM.

                Comment


                • Yeah, I think that fits. Anything that slows the driver down relative to the speed the chip consumes buffers seems to affect the amount of corruption. Roughly.
                  Test signature

                  Comment


                  • Originally posted by pingufunkybeat View Post
                    It seems to me that more complex games suffer more corruption than the simple games. Stuff like supertux is almost perfect, with a flashing triangle here or there, Neverball has more flickering, but is completely playable, while OpenArena looks like a depressed Picasso did the level design
                    can you post screenshots? maybe Picasso's soul is stuck in the chips ...

                    Comment


                    • Here is an example from OpenArena: Imageshack.

                      Still, I managed several frags, which is better than the last time I tried (when it was truly a random mess). It also doesn't speak highly of the OpenArena bots

                      The corruption in Neverball doesn't show up on screenshots. I guess the flashing triangles are too fast.
                      Last edited by pingufunkybeat; 18 August 2009, 09:05 PM.

                      Comment

                      Working...
                      X