Announcement

Collapse
No announcement yet.

The RadeonSI/RADV Performance With Mesa 19.1

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

  • The RadeonSI/RADV Performance With Mesa 19.1

    Phoronix: The RadeonSI/RADV Performance With Mesa 19.1

    If all goes well, Mesa 19.1 could debut as soon as next week. Here is a look at how the AMD Radeon Polaris and Vega performance is looking with Mesa 19.1 relative to Mesa 19.0 stable.

    http://www.phoronix.com/vr.php?view=27905

  • #2
    Is esync used in the dxvk tests?
    I guess it's not since it requires rising nofile ulimit.
    But it should be enabled for testing graphic drivers and cards.

    Comment


    • #3
      Originally posted by Etherman View Post
      Is esync used in the dxvk tests?
      I guess it's not since it requires rising nofile ulimit.
      But it should be enabled for testing graphic drivers and cards.
      Proton enabled esync by default and can be disabled by PROTON_NO_ESYNC, so no tweak has to be done by user.
      Increasing nofile ulimit no longer required on systemd 240 or later.

      Comment


      • #4
        Can you benchmark Deus Ex: MD OpenGL (native) vs DXVK ?

        Comment


        • #5
          Originally posted by Etherman View Post
          Is esync used in the dxvk tests?
          I guess it's not since it requires rising nofile ulimit.
          But it should be enabled for testing graphic drivers and cards.
          No longer necessary since systemd > 240. Michael is testing on Ubuntu 19.04 so systemd should be up to date, which means no need for increasing nofile ulimit.

          Comment


          • #6
            Ok I didn't know about that it is fixed by systemd.

            Comment


            • #7
              phoronix there's something very wrong with your F1 2018 results: you get 254 fps with an RX 590 in 1440p ultra high while the same GPU got 100fps in 1440p High (even the RTX 2070 only got 166fps in 1440p high) in your previous test: https://www.phoronix.com/scan.php?pa...gtx-1660&num=3

              Edit: even the 1080p results don't make much sense, they're way too high.

              Comment


              • #8
                Originally posted by Etherman View Post
                Ok I didn't know about that it is fixed by systemd.
                Me neither. I'm suspecting the doubled limits vs. systemd's current limits were also responsible for some alt + tabbing issues with Windows fullscreen exclusive mode vs. borderless here.

                Comment


                • #9
                  Originally posted by aufkrawall View Post
                  Me neither. I'm suspecting the doubled limits vs. systemd's current limits were also responsible for some alt + tabbing issues with Windows fullscreen exclusive mode vs. borderless here.
                  Proton should more or less force borderless AFAIK?

                  Could the small performance gain be attributed to the recent DXVK change that makes use of a newer Vulkan extension (and could slightly regress performance in other cases) ?

                  Comment


                  • #10
                    Originally posted by [email protected] View Post
                    Proton should more or less force borderless AFAIK?
                    Technically, there probably shouldn't be any difference at all between "borderless" and "exclusive" on xorg, but still there there was that weird behavior...

                    Comment

                    Working...
                    X