Announcement

Collapse
No announcement yet.

Radeon Gallium3D MSAA Performance (R300g)

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

  • phoronix
    started a topic Radeon Gallium3D MSAA Performance (R300g)

    Radeon Gallium3D MSAA Performance (R300g)

    Phoronix: Radeon Gallium3D MSAA Performance (R300g)

    This week support for MSAA was finally added to the R300g driver so that the Radeon X1000 graphics cards and earlier can finally take advantage of anti-aliasing with this open-source Gallium3D driver. In this article are some benchmarks of the MSAA performance with a Radeon X1800XT, but even with this higher-end GPU when it comes to the R300g support coverage, the anti-aliasing performance isn't really usable.

    http://www.phoronix.com/vr.php?view=18360

  • oibaf
    replied
    The torcs test was very informal, I just took the on screen fps value after starting a game, the real gap may be smaller (or bigger). Anyway I'll see to disable it.

    Leave a comment:


  • agd5f
    replied
    Originally posted by oibaf View Post
    I suspect the slowness of Phoronix test here is due to this.
    That doesn't explain the slower results with MSAA disabled.

    Leave a comment:


  • marek
    replied
    Originally posted by oibaf View Post
    Ah, OK, so your big performance drop is due to compiling with -O0. So --enable-debug + forcing -O2 shouldn?t be much slower than without debug?
    For your PPA, I prefer not to use --enable-debug and -DDEBUG at all. -DDEBUG is for development only, there's a lot of sanity checking, which reduces performance; not as much as --enable-debug, but it's noticable. --enable-debug only gives you good gdb experience, other than that it's useless.

    You said: "torcs 7.3 with debug 6.7 fps", that's with -O2 and -DDEBUG if I understood correctly. If you convert it to percents, you get 8% worse performance with -DDEBUG - that is the cost of all the sanity checking. That's is too much. Imagine every user using your PPA gets 8% lower performance...

    Leave a comment:


  • oibaf
    replied
    Originally posted by marek View Post
    No, I tested with the defaults of --enable-debug, i.e. -O0.
    Ah, OK, so your big performance drop is due to compiling with -O0. So --enable-debug + forcing -O2 shouldn?t be much slower than without debug?

    Leave a comment:


  • marek
    replied
    No, I tested with the defaults of --enable-debug, i.e. -O0.

    Leave a comment:


  • oibaf
    replied
    Originally posted by marek View Post
    With RV730:

    glxgears - 5900 FPS, with debug 3900 FPS
    torcs - 33 FPS, with debug 14 FPS
    nexuiz - 56 FPS, with debug 54 FPS (okay that is GPU-bound with graphics options maxed out)
    On my RV530:
    • glxgears - 1750, with debug 1700 FPS
    • torcs 7.3 with debug 6.7 fps


    Note that with debug I reverted the -O0 flag and used -O2. Did you do the same to get your numbers?

    Also I am sure torcs was much faster some time ago, I got around 15 fps IIRC.

    Leave a comment:


  • marek
    replied
    The issue might be that MSAA renderbuffers are allocated in RAM instead of VRAM. Who knows.

    Leave a comment:


  • Aphax
    replied
    I noticed with r600g (radeon 5850) that MSAA performance was rather slow (~20fps) when I have my OpenGL window fullscreened, as opposed to just maximizing the window (which is only slightly smaller than fullscreen) which would give me 100+ fps. This was under XFCE with no compositing, swapbufferswait off, and vblank_mode=(whatever it was that disables vsync). Maybe that's the same (or a related) issue here with the odd MSAA numbers for r300g?

    Leave a comment:


  • oibaf
    replied
    Originally posted by curaga View Post
    No online selection for patches?
    No, AFAIK.

    Leave a comment:

Working...
X