Announcement

Collapse
No announcement yet.

Many Linux Desktop 2D Benchmarks Of NVIDIA vs. AMD Drivers

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

  • Many Linux Desktop 2D Benchmarks Of NVIDIA vs. AMD Drivers

    Phoronix: Many Linux Desktop 2D Benchmarks Of NVIDIA vs. AMD Drivers

    This week on Phoronix were many Linux graphics tests from Unreal Engine 4 to Metro Redux with twenty-two AMD Radeon and NVIDIA GeForce graphics cards. Yesterday was also a 22-way GPU Linux OpenCL comparison. For your weekend viewing pleasure are now 2D desktop benchmarks from all of these GeForce and Radeon graphics cards atop Ubuntu Linux.

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    At first I thought "oh, nice. Finally some benchmarks where AMD doesn't get their ass handed to them." But then I saw that it was mostly "less is better" benchmarks.

    I think I recall that the new AMD driver (whenever that arrive) that uses AMDGPU was supposed to use GLAMOR for the 2D stack. Maybe things will get better then. At least for 2D.

    Comment


    • #3
      I don't know about GCN, but on NI and below the OSS drivers annihilate catalyst.

      Also nvidia's proprietary driver always scores well on 2d benchmarks, but it "feels" slow on the desktop.

      Comment


      • #4
        Originally posted by duby229 View Post
        I don't know about GCN, but on NI and below the OSS drivers annihilate catalyst.

        Also nvidia's proprietary driver always scores well on 2d benchmarks, but it "feels" slow on the desktop.
        That's also true for SI and above - Gallium3D is faster than Catalyst.

        Comment


        • #5
          Originally posted by mmstick View Post
          That's also true for SI and above - Gallium3D is faster than Catalyst.
          I'm sure there will be a similar comparison on the OSS drivers soon. That's usually how it gets done. One article for the proprietary drivers, another for the OSS drivers. And then usually a third article with the results of both earlier articles.

          That's ok with me. I just would like to see the OSS driver articles first.

          Comment


          • #6
            EDIT: Although, I'm absolutely positive that if Micheal did the OSS driver articles first, people would complain. So it's all good. I don't care really.

            Comment


            • #7
              amd has always the longest bars and no not all benchmarks were lower is better based.

              Comment


              • #8
                I wonder why the performance of the AMD HD6870 was consistently worse than the 6570, the former is a much better card. In some cases even the 6450 beat the 6870. Weird.

                Comment


                • #9
                  IIRC that's usually for some indirect reason, eg desktop compositing keeps the small GPU busy (and hence clocks up) while the big chip can composite with idle clocks.

                  Some benchmarks (and I'm guessing the 2D ones are good candidates) don't put enough average load on the GPU to make it look busy and have the clocks go up as a result, so the benchmark runs at whatever clocks are indicated by the background activity. Just a guess though...
                  Test signature

                  Comment


                  • #10
                    Originally posted by bridgman View Post
                    IIRC that's usually for some indirect reason, eg desktop compositing keeps the small GPU busy (and hence clocks up) while the big chip can composite with idle clocks.

                    Some benchmarks (and I'm guessing the 2D ones are good candidates) don't put enough average load on the GPU to make it look busy and have the clocks go up as a result, so the benchmark runs at whatever clocks are indicated by the background activity. Just a guess though...
                    Assuming your guess is correct, what can be done to force the GPU clock up that wouldn't interfere with benchmark results?

                    Comment

                    Working...
                    X