Announcement

Collapse
No announcement yet.

Java Performance: Ubuntu Linux vs. Windows Vista

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

  • #11
    Originally posted by sc3252 View Post
    +1 to this.
    I would also like to know if it was a clean slate from benchmarking, I understand if you couldnt because of time, but it would be nice to know.
    Each time the OS took up the entire disk.
    Michael Larabel
    http://www.michaellarabel.com/

    Comment


    • #12
      Originally posted by ethana2 View Post
      page 2

      Sunflow, in seconds
      'less is better'
      Ubuntu : 19.65, 19.75
      Vista : 20.2

      'Windows Vista was faster, but not even by a few points.'

      The graph indicates the Vista was slower, I think that's a typo.
      Delete this comment when issue is resolved.
      Oops, typo. It's now been fixed, thanks.
      Michael Larabel
      http://www.michaellarabel.com/

      Comment


      • #13
        I must say, very surprising and interesting results! I remember some tests a couple of years ago where XP was a mile ahead in almost any test, so I didn't expect this!

        Comment


        • #14
          Originally posted by ProTech View Post
          Why don't you used the same version of java on both Windows and Linux. There was big changes in 6u10 (1.6.0_10). e.g.: use Direc3D to speed up rendering on Windows. So it would been better to use that version on Windows not just on Linux. Probably Linux would beat Windows with that version too.
          I completely agree with you. This seems a typical Phoronix-test to me. The test results can't be compared if you don't use an identical version of Java.

          And why does Phoronix doesn't use the same hardware for all (or at least for most) of its tests? One time it's a AMD Phenom, another time it's a Core2Duo and so on.

          Comment


          • #15
            I do believe the comparison is not so much windows vs linux but client vm vs server vm. The difference for scimark on windows is a clear indication that the client hotspot compiler was used while the server vm was used under linux.

            Can you please rerun the windows scimark benchmark with the -server flag?

            Comment


            • #16
              Why you only benchmark the 6_7 on windows against 6_10 on ubuntu? I think there are some differences between 6_7 and 6_10 on ubuntu and windows.

              Comment


              • #17
                why only ubuntu + vista?

                opensolaris is said to have the fastest java stack so far.

                Comment


                • #18
                  Thanks for this interesting article.

                  Please consider including the IBM JDK on Linux into the comparison, as it also performs very well. And a comparison of 32-bit vs. 64-bit would also be great.

                  Comment


                  • #19
                    why only ubuntu + vista?

                    opensolaris is said to have the fastest java stack so far.
                    Good point!
                    It would be a nice comparison.
                    Though I do enjoy seeing ubuntu perform better than the alternative in any case

                    Comment


                    • #20
                      Originally posted by krausest View Post
                      I do believe the comparison is not so much windows vs linux but client vm vs server vm. The difference for scimark on windows is a clear indication that the client hotspot compiler was used while the server vm was used under linux.

                      Can you please rerun the windows scimark benchmark with the -server flag?
                      Interesting point. The 64bit JVM only has a "server" vm, while the 32bit JVM defaults to using the "client" vm, which tries to load classes faster at the expense of performance. e.g. if you run azureus with a 32bit jvm, it helps performance a _lot_ to use java -server.

                      I think the Vista screenshot shows that it's 32bit. But I don't see any clues either way about whether x86 or AMD64 Ubuntu was used.

                      There are so many variables between the systems benchmarked that all it tells you is how well Java programs will do on a default install of Ubuntu Intrepid or Vista as the primary OS on the HD. Which is interesting for a variety of reasons. If I cared more about Java performance or Vista, I'd probably join everyone else in asking for more benchmarks to be added so we could tell what was causing what.

                      Comment

                      Working...
                      X