Announcement

Collapse
No announcement yet.

Unity 5.12 Fixes Ubuntu OpenGL Performance Problems

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

  • Unity 5.12 Fixes Ubuntu OpenGL Performance Problems

    Phoronix: Unity 5.12 Fixes Ubuntu OpenGL Performance Problems

    Issued as a stable release update to Ubuntu 12.04 LTS last week was Unity 5.12. Aside from offering some minor usability enhancements and various fixes, Unity 5.12 should fix some of the OpenGL performance problems that many users have experienced -- and multiple Phoronix articles have noted the OpenGL performance slowdown -- so here's some tests seeing how Unity 5.12 now affects the OpenGL gaming performance.

    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
    Nice article, and great work by the Ubuntu dev's on addressing the performance issues in Unity!

    Comment


    • #3
      There are still some performance issues with Unity. Compared to a standalone Compiz, Unity has some performance problems with scrolling in applications (e.g. Firefox). Scrolling is faster and "softer" when using Compiz standalone.

      Also windowed OpenGL-applications run a bit faster (~5-10%) with Compiz. But that depends heavily on the used hardware.

      P.S.:

      <self-adulation>

      BTW, i was the guy who found the regression. I was so pissed that Unity 5.10 and 5.12 sucked so much on my Latide D830 that a bisected the Bazaar source code and tried to find the regression. And i found it

      </self-adulation>

      Comment


      • #4
        Michael, you should consider dropping the FPS scale and switch to raw milliseconds, as your results are (very often) misleading.

        World of Padman results:
        5.10.0: 99.86 FPS
        5.12.0: 104.70 FPS

        A difference of 4 FPS; quiet nice, isn't it? If the game would have been running at 19 FPS originally then it would put it in the fully playable-and-smooth category, right? Wrong.

        99.86 FPS = 10.013 ms per frame
        104.70 FPS = 9.551 ms per frame

        A difference of only 0.4619 ms. If the game originally ran at 19 FPS then this "speed improvement" would only have given us 0.1653 FPS difference.

        Also, if you would have used milliseconds per frame you would have noticed that, a) in the majority of cases it doesn't brake the ~2ms barrier, and b) the speedup of that magnitude is mostly irrelevant when it is the most needed. (At 19 FPS the speedup would be whopping 0.7 FPS, at 10 FPS it would be ~0.2 FPS.)

        Differences in milliseconds:
        • Doom 3: 2.2578 ms speedup on 5.12.0 compared to 5.10.0
        • Lightsmark: 2.0557 ms
        • Nexuiz: 4.004 ms
        • Open Arena 0.8.5 (900p): 0.8572 ms
        • Open Arena 0.8.8 (900p): 1.2660 ms
        • Open Arena 0.8.5 (1080p): 1.9571 ms (Wow, 40 FPS difference! What a performance improvement! Nope, not really. Still 2 milliseconds difference.)
        • Open Arena 0.8.8 (1080p): 0.6372 ms
        • Reaction Quake: 50.0205 ms (Here we *do* have a huge performance win, but from FPS you couldn't tell. It's only a difference of 22 FPS, compared to 40 we had with Open Arena.)
        • Prey: 56.3318 ms (Again, huge speedup. And again we have ~50ms difference. Prey most likely triggered the same codepath that Reaction Quake did than resulted in a 50ms loss on 5.10.0.)
        • Xonotic (Low): 0.1934 ms
        • World of Padman: 0.4619 ms
        Last edited by kouhei; 11 June 2012, 12:48 PM.

        Comment


        • #5
          Question for Michael

          On future benchmarks, when you have a test running on Ubuntu 12.04, does that mean the out-of-the-box unpatched version, or a fully-updated version?

          Same thing applies to Fedora, which tends to issues normal updates that bump the kernel version, etc., and other distros as well.

          Comment


          • #6
            Originally posted by smitty3268 View Post
            On future benchmarks, when you have a test running on Ubuntu 12.04, does that mean the out-of-the-box unpatched version, or a fully-updated version?

            Same thing applies to Fedora, which tends to issues normal updates that bump the kernel version, etc., and other distros as well.
            All depends on the scenario, etc. When looking at the Phoronix Test Suite table it shows you what versions of the key packages are in use, so from there you know if the system is updated or not... but in terms of Ubuntu benchmarking, most of my interest has already turned to 12.10 due to newer kernel and GCC.
            Michael Larabel
            https://www.michaellarabel.com/

            Comment


            • #7
              Hmm, I'm wondering if this update is somehow why Google Earth now crashes on me. I had no issues with zoom-crashes before last week, and now I get a full-on lock up. Searches suggest a bug in i915 openGL drivers, but like I said, it used to work just fine. I did install an SSD as well, but everything else has been very stable. I wish Google would work on the Linux version of Earth more.

              Comment


              • #8
                Too bad they didn't fix the "removal of dodge windows" regression that was introduced recently...

                Comment


                • #9
                  Thanks for glasen contributing his time to find the issue, the Ubuntu team for fixing it and Michael for letting me know it's fixed

                  Comment


                  • #10
                    [Test Case] Performance issue - just checking if performance in OpenGL applications with Unity loaded are normal. [Regression Potential] Visual regressions, shell drawing problems. Part of a big change, many regression potentials. Original description: The main problem is that OpenGL apps run slower with unityshell compared to plain compiz (Gnome Classic). In extreme cases, running glmark2 under Unity only achieves 0-3 FPS. Yet on the same machine reports frame rates of several hundred FP...

                    Comment

                    Working...
                    X