Announcement

Collapse
No announcement yet.

Super Pi : Linux vs Windows ?

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

  • Super Pi : Linux vs Windows ?

    Ran Super PI in Linux quite a few times and the results are pretty different.

    Here is my best yet in Windows :P



    Here are the scores for Linux



    Phoronix Global - TE Testing


    Why there is so much difference ? :S

    Also my system boot with X4 B50 running @ 3.9+ghz , i know its not stable but in windows it didn't even boot :P

    The system [X4 b50] is stable 24x7 @ 3.82, Stressed and tested :P :P

    Please clear the doubts ..

    Regards

  • #2
    - Better memory management
    - linux is running at max clocks by default while windows puts cpu to lower clocks and raises the clocks after a moment
    - Better optimised build for linux
    - Windows has more background activity taking time
    - Multi-threaded? Better scheduler/threading

    So many small stuff that can affect a lot in end results.

    Comment


    • #3
      Originally posted by Dark_Star View Post
      Why there is so much difference ? :S
      these are different programs, you can't compare the results
      run the Windows SuperPi 1.5 through WINE and then compare, but that result is wrong too, due to WINEs' overhead anyway

      Comment


      • #4
        1)Super Pi is Super Pi no matter which language it is compiled under.
        It's the same program, performing the same functions.

        2) Linux has always been much more efficient than Windows.

        3) Super-Pi is a 32 bit binary so the multi-threading benefits are very limited.

        4) Back to the original question: Windows runs a ton of applications and processes that consume extreme amounts of memory and processor cycles. Linux is much cleaner and leaner in this regard. Run comparative benchmarks on the same hardware for just about any process you can imagine and Linux will out perform Windows every time. From calculating Pi to converting audio, to converting video and even running games.

        Comment


        • #5
          Originally posted by Jaguar07 View Post
          Run comparative benchmarks on the same hardware for just about any process you can imagine and Linux will out perform Windows every time. From calculating Pi to converting audio, to converting video and even running games.
          And yet this doesn't seem to be so.

          http://www.anandtech.com/systems/sho...px?i=3616&p=24

          Comment


          • #6
            Now for further debunkng

            Originally posted by Jaguar07 View Post
            1)Super Pi is Super Pi no matter which language it is compiled under.
            It's the same program, performing the same functions.
            False, different languages often use different analyzers, intermediate code generators and intermediate code optimizers. Depending how all those interpret your language can make for vastly different binaries at the machine code level.

            2) Linux has always been much more efficient than Windows.
            Again read above link, it's simply not true.

            3) Super-Pi is a 32 bit binary so the multi-threading benefits are very limited.
            Well that's just a crock of crap. Being 32 or 64 bit has no effect on multithreading performance for the most part.

            If you perceive a difference chances are that the 32-bit binary was compiled with less special instruction sets that are typically found by default on a 64-bit system (ie SSE 2). Even then if the 32-bit binary was compiled with the same support most of the time the difference is non existent.
            Last edited by deanjo; 09-11-2009, 08:53 AM.

            Comment


            • #7
              Originally posted by deanjo View Post
              And yet this doesn't seem to be so.

              http://www.anandtech.com/systems/sho...px?i=3616&p=24
              Meaningless - there's no any data and test specification. I suppose Jaguar07 was talking about real world benchmarks and it's very possible he's right.

              You gave something like this:

              http://blog.ibeentoubuntu.com/2009/0...ws-7-beta.html

              but here we see at least some apps version (they're different, so also meaningless ).
              Last edited by kraftman; 09-11-2009, 09:32 AM.

              Comment


              • #8
                Originally posted by deanjo View Post
                And yet this doesn't seem to be so.

                http://www.anandtech.com/systems/sho...px?i=3616&p=24
                Comparisons of this type are meaningless due to the enormous number of variables affecting the result besides the OS.

                Linux or Windows can only ever be "best" for a specific application, using a specific version of a specific software stack compiled using a specific compiler - nothing more.

                edit: beaten by kraftman

                Comment


                • #9
                  Originally posted by krazy View Post
                  Comparisons of this type are meaningless due to the enormous number of variables affecting the result besides the OS.

                  Linux or Windows can only ever be "best" for a specific application, using a specific version of a specific software stack compiled using a specific compiler - nothing more.

                  edit: beaten by kraftman

                  That's right. It seems only some third party members play in such benchmarks to bring some attention ;> Many are just meaningless like this one:

                  http://lkml.indiana.edu/hypermail/li...04.1/1571.html

                  I'm fast

                  Comment


                  • #10
                    Originally posted by krazy View Post
                    Comparisons of this type are meaningless due to the enormous number of variables affecting the result besides the OS.

                    Linux or Windows can only ever be "best" for a specific application, using a specific version of a specific software stack compiled using a specific compiler - nothing more.

                    edit: beaten by kraftman
                    Never the less it still proves the statement:

                    Run comparative benchmarks on the same hardware for just about any process you can imagine and Linux will out perform Windows every time. From calculating Pi to converting audio, to converting video and even running games.
                    completely false and unfounded.

                    Comment


                    • #11
                      Indeed I am referring to actual benchmarks as opposed to synthetic.

                      I would also like someone, preferably, deanjo to find an application for Linux or compatible with Windows7 that is a 32 bit application that utilizes more than 2 processors simultaneously. I do not believe you will have much success in that regard. If memory serves me, I do not believe you would be able to find any applications for windows that will use more than 2 processors simultaneously that are 32 bit applications. If you find one let me know please. I'll be happy to test that it can use more than 2 processor cores simultaneously and correct my earlier post related to 32bit vs 64bit multi threading capabilities.

                      64 bit is an entirely different matter. There are many applications written for 64 bit that have correct multi threading capabilities. Though not nearly as many as I would hope there to be at this stage of 64 bit hardware availability.

                      Comment


                      • #12
                        Originally posted by Jaguar07 View Post
                        Indeed I am referring to actual benchmarks as opposed to synthetic.

                        I would also like someone, preferably, deanjo to find an application for Linux or compatible with Windows7 that is a 32 bit application that utilizes more than 2 processors simultaneously. I do not believe you will have much success in that regard. If memory serves me, I do not believe you would be able to find any applications for windows that will use more than 2 processors simultaneously that are 32 bit applications. If you find one let me know please. I'll be happy to test that it can use more than 2 processor cores simultaneously and correct my earlier post related to 32bit vs 64bit multi threading capabilities.

                        64 bit is an entirely different matter. There are many applications written for 64 bit that have correct multi threading capabilities. Though not nearly as many as I would hope there to be at this stage of 64 bit hardware availability.
                        LMAO seriously, umm lets try 3ds max, maya, tones of encoding apps, photoshop, divx, lame-mt, etc etc etc.

                        32-bit has nothing to do with being multithreaded.

                        Comment


                        • #13
                          More winrar, 7-zip, mysql, and here is a small list found here:

                          http://www.denniskarlsson.com/smp/

                          Comment


                          • #14
                            Yes and no. 32 bit in and of itself does not require multi threading to be turned off. Mostly it has to do with the lack of backwards revision. Most organizations and people that create code do not go back and update the 32 bit version when they are only going forward with the 64 bit version. So if they were working on 64 bit code when they learned how to make their code multi thread for more than 2 cores, they typically don't go back and update the 32 bit version.

                            Maya 32 bit does not use more than 2 cores simultaneously. I do not have a copy of 3ds max installed to test with. Photoshop 32 bit does not use more than 2 cores in Windows 7 Ultimate 64 bit or XP Pro 64 bit. DivX players do not use more than 2 cores when 32 bit on Windows 7 Ultimate 64 bit, XP Pro 64 bit. Not sure about the licensed compiler version of DivX. I do know the 64 bit Beta for Adobe Flash plugin does not use more than 2 cores simultaneously.

                            So in theory you are accurate that just because something is 32 bit or 64 bit does not indicate it will or will not multi-thread correctly.

                            Where the rubber meets the road, find a 32 bit application that I can download and test, preferably part of the standard linux distro, that can use more than 2 cores at once.

                            Super-pi in the rendition of Phoronix-Test-Suite does not use 2 cores simultaneously. It does alternate between 2 different cores, but uses only 1 core at a time. Cores 0 and 2 were untouched. Cores 1 and 3 were alternated.

                            Comment


                            • #15
                              I'll test a few of those and see if I get results that work with more than 2 cores simultaneously.

                              Comment

                              Working...
                              X