Announcement

Collapse
No announcement yet.

Ubuntu 7.04 to 8.10 Benchmarks: Is Ubuntu Getting Slower?

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

  • Ubuntu 7.04 to 8.10 Benchmarks: Is Ubuntu Getting Slower?

    Phoronix: Ubuntu 7.04 to 8.10 Benchmarks: Is Ubuntu Getting Slower?

    With the release of Ubuntu 8.10 coming out later this week we decided to use this opportunity to explore how the performance of this desktop Linux operating system has evolved over the past few releases. We performed clean installations of Ubuntu 7.04, Ubuntu 7.10, Ubuntu 8.04, and Ubuntu 8.10 on a Lenovo ThinkPad T60 notebook and used the Phoronix Test Suite to run 35 tests on each release that covered nine different areas of the system. After spending well more than 100 hours running these tests, the results are now available and our findings may very well surprise you.

    http://www.phoronix.com/vr.php?view=13022

  • #2
    Greetings Phoronix-team!

    Do you plan to extend that benchmark-testing to other hardware-setups and/or other distributions?

    Best regards ...

    MacSlow

    Comment


    • #3
      Hi MacSlow,

      Yeah there's plans to at least run the same set of tests on Fedora on the T60. Possibly other distros too depending upon time and what the results are from Fedora. Other hardware setups too are planned. Ideally the Fedora results would be out by week's end.

      Best,
      Michael
      Michael Larabel
      http://www.michaellarabel.com/

      Comment


      • #4
        I believe that canonical have a large set of machines that run automated tests of development releases. Maybe you could talk them into including performance tests.

        The sooner the regressions are spotted, the easier to find which change caused them. The dramatic change in RAM speed must surely be some kernel bug.

        Have you tried 64 bit tests? A lot of the optimisation in GCC is not enabled in 32bit ubuntu to keep compatibility with i586.

        Comment


        • #5
          Cool work on these tests.

          It looks like these might be kernel regressions. You can try the kernels from the older distros on the newer ones to check if that's the case.

          Also, trying out the 64-bit distros helps. As ssam noted, the kernels and packages are not compiled with several optimizations enabled. Also, most of the newer distros will be installed on newer hardware, which means it's 64-bit capable. There's really no reason to run 32-bit distros these days.

          Comment


          • #6
            Interesting results.
            Phoroix should also do a similar tests on some server(http, db, file etc) versions (both between releases and distros if possible). There will be lot of interest regarding this aspect.

            And ofcourse, great work. These are the reviews which we actually need.

            Comment


            • #7
              that is no problem of canonical...it's a problem that affects the kernels...since 2.24 the kernels are getting slower and slower...I've heared that from sidux users, from fedora users and from Debian users...they are all saying the same: since 2.24 they are loosing frames in first-person-shooter like ET:QW and UT2k4

              so, maybe someone should talk to the kernel hackers and say them, that a sleep function after each function is NOT necessary...

              Comment


              • #8
                It might be possible to locate these regressions by running a newer backported kernel on ubuntu 7.04 (or possibly run the older 2.6.20 kernel with 8.10).

                This could isolate the issue to the kernel, or otherwise to another component, such as GCC. From there, it should be possible to further bisect the kernel to find the source of these performance regressions.

                Keep up the excellent work!

                Comment


                • #9
                  Eh, it didn't look like such a huge loss to me overall :-/

                  Comment


                  • #10
                    The cause for the bad performance is the CFS (Completly Fair Scheduler), I think. It's standard since 2.6.23/24.

                    World of Warcraft on Wine lags terrible and freezes every some seconds since the introducing of CFS.

                    I hope OpenSolaris supports my hardware completly soon, so that I can change to this OS.

                    Comment


                    • #11
                      Evgeniy Polyakov has made some posts on kernel performance. maybe a similar cause.
                      http://tservice.net.ru/~s0mbre/blog//devel/other

                      Comment


                      • #12
                        ...

                        Since there is now a complete shift to dual cores as a standard, and Intel Core Duo is not really dual cores, I wonder how the tests would perform under that environment.

                        Comment


                        • #13
                          Originally posted by trooper09 View Post
                          Since there is now a complete shift to dual cores as a standard, and Intel Core Duo is not really dual cores, I wonder how the tests would perform under that environment.
                          intel's naming scheme is confusing

                          core solo = 32bit single core
                          core duo = 32bit dual core
                          core 2 solo = 64 bit single core
                          core 2 duo = 64 bit dual core
                          core 2 quad = 64 bit quad core
                          core 2 extreme = 64 bit dual/quad

                          the dual cores share cache (but that is a good thing (i think. lets you do openMP))

                          The quads are actually two bits of silicon in one enclosure.

                          there is far too much information on wikipedia :-)

                          Comment


                          • #14
                            I was wondering if Tracker was enabled or disabled in the later versions of Ubuntu. One thing I noticed right away in Ubuntu 7.10 when I started compiling larger projects is that there was a huge performance hit taken with Tracker enabled. I would download a tarball, decompress it, and tracker would index all of these new text files... not good, especially for a large project. Then, as I was compiling, Tracker would notice these new files created during compilation and start indexing them as well. Disabling Tracker completely helped general performance quite a bit as well and it's one of the first things I do when I perform a fresh install of Ubuntu.

                            When it comes to comments on the completely fair scheduler, there may be a small performance hit taken, but the general user experience is much more smooth on the desktop. If there is a single application that runs a muck, it's not going to make your desktop completely unusable to the point of stopping the user from killing the wild process. For me, that's a trade off I can accept.

                            For the graphics tests (minus the games tests), I would be interested to see what the performance is like if we take the proprietary ATI drivers out of the mix. I realize that many (if not most) users will be using the proprietary drivers, but I'm curious to see if some of these slow downs have been introduced by ATI.

                            I'm not denying for one second that there is a problem here (it seems pretty apparent there's a problem with the kernel, X, GTK, or some combination of those), but these are just a few questions and opinions about what the causes might be.

                            Comment


                            • #15
                              Originally posted by thacrazze View Post
                              The cause for the bad performance is the CFS (Completly Fair Scheduler), I think. It's standard since 2.6.23/24.

                              World of Warcraft on Wine lags terrible and freezes every some seconds since the introducing of CFS.

                              I hope OpenSolaris supports my hardware completly soon, so that I can change to this OS.
                              That's a wicked thought. The CFS was created to improve desktop performance, particularly for things like 3d games. Maybe Con Kolivas was right all along...

                              Michael, what you should try to do is boot a Feisty kernel in Intrepid and then benchmark that (if possible).

                              Comment

                              Working...
                              X