Announcement

Collapse
No announcement yet.

More Mesa 10.2 + Linux 3.14 Benchmarks On Ubuntu 14.04 LTS

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

  • More Mesa 10.2 + Linux 3.14 Benchmarks On Ubuntu 14.04 LTS

    Phoronix: More Mesa 10.2 + Linux 3.14 Benchmarks On Ubuntu 14.04 LTS

    In verifying and complementing the major R600/RadeonSI performance regressions noted earlier today for a wide-range of AMD Radeon hardware, here's some more tests that were done this weekend on a different system that also finds the newer code to be running slower than Linux 3.13 + Mesa 10.1..

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

  • #2
    good news about improving. I hope also tests by HD3850 AGP... that's my card.

    Comment


    • #3
      However, this doesn't appear to be the only cause for the performance drop when upgrading Ubuntu 14.04's graphics stack. Other regressions have also been reported by other users of the open-source AMD driver. "This is not completely from HyperZ and so still worth investigating. We have reports from Luke and dungeon on this forum that confirm there's another regression besides the intended hyperz change. Luke specifically tested it, dungeon's case is media apps that do not use the Z buffer." There's more comments in this forum post about the issue unrelated to HyperZ.
      So for my r200 regression

      bad
      https://git.kernel.org/cgit/linux/ke.../next-20140228

      bad
      https://git.kernel.org/cgit/linux/ke...=next-20140115

      good
      https://git.kernel.org/cgit/linux/ke...=next-20140114

      and then:

      good
      http://cgit.freedesktop.org/~airlied...3e2c930f3f5706

      BINGO
      bad
      http://cgit.freedesktop.org/~airlied...a3867867216103
      Last edited by dungeon; 03-02-2014, 10:01 PM.

      Comment


      • #4
        That is enough for today . Maybe i will checked agd5f tree further for the right commit tomorrow

        Comment


        • #5
          Found it . My bad it is not in radeon, but change in ttm... this commit:

          http://cgit.freedesktop.org/~airlied...fda6ef42b53eb5

          It says "Use PFNMAP instead of MIXEDMAP where possible for performance", but actually this make it running poor on r200 and consume nearly 500 MHz more CPU here .

          Comment


          • #6
            Originally posted by dungeon View Post
            Found it . My bad it is not in radeon, but change in ttm... this commit:

            http://cgit.freedesktop.org/~airlied...fda6ef42b53eb5

            It says "Use PFNMAP instead of MIXEDMAP where possible for performance", but actually this make it running poor on r200 and consume nearly 500 MHz more CPU here .
            Please file a bug (https://bugs.freedesktop.org) and cc Thomas.

            Comment


            • #7
              Originally posted by agd5f View Post
              Please file a bug (https://bugs.freedesktop.org) and cc Thomas.
              OK filled https://bugs.freedesktop.org/show_bug.cgi?id=75719

              Comment


              • #8
                I have a Mint 16 install that I use for games. I recently got sick and tired of fglrx and the nonsense I have to go through every time I want to use a newer kernel, and the performance regressions (on my hardware) in the last two fglrx betas. (I was stuck on catalyst 13.11 Beta 9.4... the last one that worked decently for me on Linux 3.12.x with my HD 5870 card)

                So I built a Linux 3.13.5 kernel, enabling the radeon/DRI options (I had them disabled because I only used fglrx), removed my fglrx debs and rebooted into a much better environment with the open source radeon driver. Desktop stuff and hardare accelerated web browsers actually work better than with fglrx. Windows create and destroy with a snap, switching between virtual desktops is snappier, scrolling is better in the browsers (Chromium and Firefox) and tab switching is instant. Of course, these some of these things don't really matter in the grand scheme of things (milliseconds don't matter to productivity), it's just a matter of perception.

                Wow... most of my games (especially source engine games) actually work better with radeon than with fglrx. I was pleasantly surprised. Left4Dead 2 (for one example) which was horrible with fglrx, was stutter free and smooth.

                Of course, that left Metro Last Light out in the cold, as it needs a newer OpenGL than the Mesa 9.2 that I had. I was going to forget about Metro LL, because while it worked decently (with the occasional crash) all the way up to that point, it was crashing every time after a load screen with fglrx and I couldn't complete it anyway.

                Then I bought Deadfall Adventures, which was crashing on the opening cut scenes (with radeon) and I couldn't even start the game. I managed to skip through them with the space bar and get to the game proper, but it was unplayable.

                So I added the xorg-edgers ppa to get Mesa 10 (10.2.0 now) and had much better luck with Deadfall Adventures (not crashing on cut scenes and playable on Medium) and Metro Last Light runs again. It still won't get past that point where it was crashing after a load screen, but at least it runs again, and every bit as well too.

                However, I have noticed a regression on the source engine games. A bit of audio and video stuttering after a level loads but mostly fine. Left 4 Dead 2 has a bit of occasional stuttering in game, but still way better than before. Noticeable though, Mesa 9.2 seemed better for those games. Oh well, a necessary evil for now. I'm sure this will improve soon enough.

                Comment


                • #9
                  Originally posted by Grogan View Post
                  [...]
                  What about graphics card noise? I heard bad reports about noise with the radeon driver.

                  Comment


                  • #10
                    Originally posted by Calinou View Post
                    What about graphics card noise? I heard bad reports about noise with the radeon driver.
                    I haven't noticed any extra noise. It's possible the fan may be running more, as the card does seem to run pretty cool (which was one of my concerns, that kept me from ditching fglrx a bit longer) but I don't notice it. I've got a lot of case fans anyway, and a heavy duty bathroom type exhaust fan in the ceiling of my office/computer room that runs 24/7. The sound of the graphics card doesn't come through over the existing white noise, so I'm going to say no on that issue.

                    Comment

                    Working...
                    X