Announcement

Collapse
No announcement yet.

Article on graphics driver - need inputs

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

  • Article on graphics driver - need inputs

    Hello guys

    I am writing an article on the state of graphics driver on linux. What I aim to do is compare Intel, AMD and Nvdia graphics support in both open source and closed alternatives.
    Till now this is how I plan to test . Use Ubuntu and Mandirva / Fedora 16 for the test environment. and test the graphics capabilities in multiple desktop environment.

    1. check desktop env, compatibility and performance. Gnome3, legacy gnome, unity and kde.
    2. glxgears as the benchmarks ?
    3. unigine benchmarks for comparing opensource and closed performance diff.
    4. 1080p and hardware acceleration.

    So what I want is you guys to suggest me some better test if available and please post your experience with performance and specially h.w acceleration in Intel and nvidia.

    This is probably the first time I am asking for inputs but I need it to make the article comprehensive. Any inputs and suggestions are welcome.

    Please do help me making a decent writeup.

    Regards

  • #2
    I disagree, unigine is a good test, as long as you enable all the patented options and install the s3tc lib.

    Comment


    • #3
      Originally posted by Qaridarium
      a 'test' is never serious only real applications and real games are serious.
      Thanks for the inputs . But then a benchmark does give an out look on the performance front of the h.w

      Anyway can you guys post some benchmarks score specially with intel and nv hardware . ?

      that will be great help Specially inputs on h.w acceleration, quick sync and nv h.w acceleration.

      Comment


      • #4
        Ok I'll keep your suggestion in mind but how do I measure fps in linux /

        Also can anyone of you please post an quick info about the support for hardware accelaration of Intel and nvidia. Specially while rendering 1080p videos, with cpu usage.

        That will be really h

        Comment


        • #5
          Originally posted by Qaridarium
          it makes a big different if you benchmark with 3dmark2006 or with a real game.
          the practical relevance of 3Dmark2006 is zero.
          same with unigine benchmarks like "heaven" practical relevance is zero.

          if you wanna do a unigine based benchmark use the game "oil-rush" and not the benchmarks.
          Oh wow.... This may be your first post that I completely agree with...

          EDIT: To the OP. FPS is reasonable to measure if you do it on a line graph, where time is compared against FPS. Take a sample of FPS every 1 second or 5 seconds, and then put it on a line graph. The more samples you take, the more meaningful it will be. Not in benchmark mode, but rather as your playing through the game. This way you can get a visual representation of how well the game actually performs as it is being played through.

          The problem with this method is that because youre playing through the game results between cards wont be exactly comparable. They will be real life, but not comparable. The reason it wont be comparable is because each run through, the frames being rendered at each individual sample wont be exactly the same.... Still I would rather have a real life benchmark, than a comparable benchmark.
          Last edited by duby229; 08 December 2011, 03:56 PM.

          Comment


          • #6
            EDIT2: The same can also be true of video playback, except this time you'll do time versus cpu usage on a line graph. Except that with video playback the results will be both real life AND comparable, because the frames being rendered at each sample will always be exactly the same.

            Comment

            Working...
            X