Announcement

Collapse
No announcement yet.

Radeon RX 580: AMDGPU-PRO vs. DRM-Next + Mesa 17.2-dev

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

  • Radeon RX 580: AMDGPU-PRO vs. DRM-Next + Mesa 17.2-dev

    Phoronix: Radeon RX 580: AMDGPU-PRO vs. DRM-Next + Mesa 17.2-dev

    Last week I posted initial Radeon RX 580 Linux benchmarks and even AMDGPU overclocking results. That initial testing of this "Polaris Evolved" hardware was done with the fully-open Radeon driver stack that most Linux enthusiasts/gamers use these days. The AMDGPU-PRO driver wasn't tested for those initial articles as it seems to have a diminishing user-base and largely focused for workstation users. But for those wondering how AMDGPU-PRO runs with the Radeon RX 580, here are some comparison results to DRM-Next code for Linux 4.12 and Mesa 17.2-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
    I find it hard to believe that the open source community (I guess with the help of paid AMD developers) can develop a driver that performs better than what AMD are doing in-house. Wow.

    Comment


    • #3
      Originally posted by molecule-eye View Post
      I find it hard to believe that the open source community (I guess with the help of paid AMD developers) can develop a driver that performs better than what AMD are doing in-house. Wow.
      That is kinda unfair. Most of the special sauce for the kernel driver came from AMD.

      Comment


      • #4
        Originally posted by molecule-eye View Post
        I find it hard to believe that the open source community (I guess with the help of paid AMD developers) can develop a driver that performs better than what AMD are doing in-house. Wow.
        As Drago said, most of the freedom driver is developed in-house by AMD. The blob probably supports some more features, esp. things in the context of DC (ex-DAL) related features, may be optimized for specific settings or applications and iirc. it's faster in terms of Vulkan: Vulkan RadV is actually not a direct AMD development and rather mainly done by AirlieD (RH) and team. AMD wanted to put their own Vulkan to freedom but haven't done that yet.
        Stop TCPA, stupid software patents and corrupt politicians!

        Comment


        • #5
          Originally posted by molecule-eye View Post
          I find it hard to believe that the open source community (I guess with the help of paid AMD developers)
          you've got it backwards. it is paid amd developers with the help of open source community

          Comment


          • #6
            Amazing results in any case

            Comment


            • #7
              Next up: AMD pushes open source OpenCL implementation to Linux? Pretty please...

              Comment


              • #8
                Wow, really impressive results! We just need to catch up with the performance of the nVidia blob next. I hardly ever boot into Windows these days, I used to reboot for stubborn games, but with so many great games on Steam and World of Warcraft playing amazing under wine with the nine state tracker - why bother?

                Comment


                • #9
                  Originally posted by pal666 View Post
                  you've got it backwards. it is paid amd developers with the help of open source community
                  AMD couldn't get such performance without the Community. What's the catch? Linux design gave such advantage, perhaps?

                  Comment


                  • #10
                    One aspect that probably helped us is that we've been comparing Mesa and our closed Vulkan driver code and porting some of the good stuff to Mesa.

                    Comment

                    Working...
                    X