Announcement

Collapse
No announcement yet.

R600g/RadeonSI Gallium3D Performance At The End Of 2017

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

  • R600g/RadeonSI Gallium3D Performance At The End Of 2017

    Phoronix: R600g/RadeonSI Gallium3D Performance At The End Of 2017

    One of the common test requests to come in for our end-of-year benchmarking has been a fresh look at the Radeon GPU performance incorporating some both old and new GPUs to see the current state of the open-source driver stack. Tests were done from a Radeon HD 5830 on the Radeon+R600g driver stack to the RX Vega 64 on AMDGPU+RadeonSI, while using the Linux 4.15-rc5 kernel paired with Mesa 17.4-dev.

    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
    Wow, Fury doing very well against the 14nm GPU's even in power efficiency. Bought my Sapphire Nitro R9 Fury on sale just after the Polaris cards came out, and for the same price as an RX480 it really feels like it was the right choice. I have a feeling I'll stick with this card for a long time. I expect Vega to get better with time though.

    Comment


    • #3
      Thanks for benchmarks Michael.

      From what I can test on my machine with HD5850 I see that my old fglrx (catalyst 15.9) driver in many games perform better than open source Radeon + mesa 17.2/4. Don't know why. I still use Mageia 5 distro with fglrx because on any distro with Radeon and Mesa, my GPU fan working at 100% speed and noise is unbearable...

      Reported bug here https://bugs.freedesktop.org/show_bug.cgi?id=101987 but looks like nobody care about fixing bugs for old gpu :/ Maybe anyone have similar issue and know how to fix it?
      Last edited by xpris; 26 December 2017, 11:05 AM.

      Comment


      • #4
        @Michael
        One thing I don't like with your benchmarks is that you use non final, non official third party components. These are useless IMO. I am interested to what we can get by using the components that are stable and officially distributed by the OS or hardware manufacturer. Also at the default settings with no customizations whatsoever. Please encourage the relevant people to find procedures such that the driver components to be easily consumed by regular users.

        Comment


        • #5
          Originally posted by zoomblab View Post
          @Michael
          One thing I don't like with your benchmarks is that you use non final, non official third party components. These are useless IMO. I am interested to what we can get by using the components that are stable and officially distributed by the OS or hardware manufacturer. Also at the default settings with no customizations whatsoever. Please encourage the relevant people to find procedures such that the driver components to be easily consumed by regular users.
          I believe the point is that this code is changing rapidly, so using the latest git pulls gives a "bleeding edge" performance snapshot. That way we can see where the development is heading, without waiting another year for it to be marked stable and adopted by the distros.

          Comment


          • #6
            Originally posted by zoomblab View Post
            @Michael
            One thing I don't like with your benchmarks is that you use non final, non official third party components. These are useless IMO. I am interested to what we can get by using the components that are stable and officially distributed by the OS or hardware manufacturer. Also at the default settings with no customizations whatsoever. Please encourage the relevant people to find procedures such that the driver components to be easily consumed by regular users.
            This is an article about how performance on the bleeding edge is, if you want articles about how it is in various distros, he makes them for Ubuntu (the officially supported distro for Steam) already.
            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


            Comment


            • #7
              Originally posted by xpris View Post
              From what I can test on my machine with HD5850 I see that my old fglrx (catalyst 15.9) driver in many games perform better than open source Radeon + mesa 17.2/4. Don't know why. I still use Mageia 5 distro with fglrx because on any distro with Radeon and Mesa, my GPU fan working at 100% speed and noise is unbearable...
              I think I remember reading somewhere (sometime in the last 6 months IIRC) that the sb optimizer had to be disabled in order to enable higher OpenGL levels, at least without a lot of additional work. If that is true, it might explain why you aren't seeing fglrx-level performance today.
              Test signature

              Comment


              • #8
                Originally posted by bridgman View Post

                I think I remember reading somewhere (sometime in the last 6 months IIRC) that the sb optimizer had to be disabled in order to enable higher OpenGL levels, at least without a lot of additional work. If that is true, it might explain why you aren't seeing fglrx-level performance today.
                Good to know, thanks. Maybe you know if is any hope for fixing fan speed on my hd5850?

                Comment


                • #9
                  Wonderful results as usual for the OpenGL drivers. And thanks for running SuperTuxKart as well. Some games might not be the most famous ones but it's interesting to see different results and mostly those of native games that haven't had to be ported.

                  Comment


                  • #10
                    It seems Tonga (285/380/480) has fallen out of the pool of tested cards. That is a shame as I believe Tonga was a very popular mid-range card.

                    Comment

                    Working...
                    X