Page 1 of 2 12 LastLast
Results 1 to 10 of 12

Thread: Gallium3D LLVMpipe Compared To Nine Graphics Cards

  1. #1
    Join Date
    Jan 2007
    Posts
    14,901

    Default Gallium3D LLVMpipe Compared To Nine Graphics Cards

    Phoronix: Gallium3D LLVMpipe Compared To Nine Graphics Cards

    Yesterday after publishing the 15-way open-source vs. closed-source NVIDIA/AMD Linux graphics comparison there were some requests by Phoronix readers to also show how the LLVMpipe software rasterizer performance is in reference. For this article to end out the month are the OpenGL performance results from nine lower-end AMD Radeon and NVIDIA GeForce graphics cards running with their respective Mesa/Gallium3D drivers compared to the LLVMpipe software driver in two configurations.

    http://www.phoronix.com/vr.php?view=18692

  2. #2
    Join Date
    Oct 2009
    Posts
    2,118

    Default

    Quote Originally Posted by phoronix View Post
    Phoronix: Gallium3D LLVMpipe Compared To Nine Graphics Cards

    Yesterday after publishing the 15-way open-source vs. closed-source NVIDIA/AMD Linux graphics comparison there were some requests by Phoronix readers to also show how the LLVMpipe software rasterizer performance is in reference. For this article to end out the month are the OpenGL performance results from nine lower-end AMD Radeon and NVIDIA GeForce graphics cards running with their respective Mesa/Gallium3D drivers compared to the LLVMpipe software driver in two configurations.

    http://www.phoronix.com/vr.php?view=18692
    And once again, a totally pointless test.

    1) Benchmark it doing something that it would actually be useful doing, like DESKTOP COMPOSITING.
    2) Benchmark it on a system typical of those that don't already have decent GPU's, like... intel Z520.

    Nobody cares how fast it can play games on a very fast 8-core processor.

  3. #3
    Join Date
    Nov 2012
    Posts
    146

    Default

    LOL that's lame. Stop complaining, do the tests yourself...
    Quote Originally Posted by droidhacker View Post
    Nobody cares how fast it can play games on a very fast 8-core processor.
    exactly that was requested:
    Quote Originally Posted by chithanh View Post
    Also I would have liked to see llvmpipe in this comparison, as the 8350's 8 cores could give results close to the low-end cards.

  4. #4
    Join Date
    Oct 2007
    Posts
    1,277

    Default

    What's up with the RadeonHD 6450 results?

  5. #5
    Join Date
    Oct 2008
    Posts
    3,149

    Default Not so bad, actually

    Getting 5-15 fps at 1080p resolution is better than i expected to see from a software renderer.

  6. #6
    Join Date
    Nov 2012
    Location
    France
    Posts
    582

    Default

    Quote Originally Posted by droidhacker View Post
    Nobody cares how fast it can play games on a very fast 8-core processor.
    4 modules actually, but you still 8 cores. You get the performance of a quad core Intel (in multithread usage), similar to a 2600(K).

  7. #7
    Join Date
    Mar 2012
    Posts
    52

    Default

    And with a weaker cpu and a low resolution display (like 1366x768, 1280x1024 ecc.)? Does it (down)scale lineary?

  8. #8

    Default

    Quote Originally Posted by droidhacker View Post
    And once again, a totally pointless test.

    1) Benchmark it doing something that it would actually be useful doing, like DESKTOP COMPOSITING.
    2) Benchmark it on a system typical of those that don't already have decent GPU's, like... intel Z520.

    Nobody cares how fast it can play games on a very fast 8-core processor.
    Well if it can't run OpenErena 8.5 at more then a handful of FPS even on a very fast CPU then any CPU that you have that would not have a supported GPU on the mobo would run compositing over LLVM at around 1 frame every 4 seconds or less.

    LLVM is not a substitute for a GPU in any case save for doing single frame rending accuracy tests.

  9. #9
    Join Date
    Jun 2010
    Location
    ฿ 16LDJ6Hrd1oN3nCoFL7BypHSEYL84ca1JR
    Posts
    1,052

    Default

    Quote Originally Posted by Kivada View Post
    LLVM is not a substitute for a GPU in any case save for doing single frame rending accuracy tests.
    Well, without knowing that much about OpenGL rendering I'm still interested in how it would perform on the Parallella with 64 of those cores: http://www.adapteva.com/products/sil...vices/e64g401/ and whether the latency could be acceptable. (with necessary modifications of course)

    The performance would probably not be too impressive, but it could still be ok. Also, 2014 they want to reach for ~1.2 TFlops

  10. #10

    Default

    Quote Originally Posted by ChrisXY View Post
    Well, without knowing that much about OpenGL rendering I'm still interested in how it would perform on the Parallella with 64 of those cores: http://www.adapteva.com/products/sil...vices/e64g401/ and whether the latency could be acceptable. (with necessary modifications of course)

    The performance would probably not be too impressive, but it could still be ok. Also, 2014 they want to reach for ~1.2 TFlops
    Likely wont be that great since they will hit the same wall that Intel did with Larrabee. It might be decent for ray tracing, but probably be terrible for standard graphics and will draw much more power then a dedicated GPU would to do the same task.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •