Crunching times on UltraSparc II

Forums › Cruncher's Corner

Dotsch
Dotsch
Joined: 1 May 05
Posts: 50
Credit: 485389
RAC: 396
23 Apr 2006 22:35:01 UTC
Topic 191128
(moderation:
    )

    I wan't to crunch einstein@home with my Ultra 30 (with a Ultra Sparc II 300 MHz CPU). How long are the typical crunching times for this CPU ?

    I have attached to the project, but the estimated runtime is after some hours about 100 hours. This is a litte bit difficult to complete a work unit within the deadline, because this system is not to often running...

    DanNeely
    DanNeely
    Joined: 4 Sep 05
    Posts: 1364
    Credit: 3562358667
    RAC: 109

    Crunching times on UltraSparc II

    23 Apr 2006 22:52:48 UTC
    Message 29320
    (moderation:
      )

      For a system that old the runtime doesn't seem unreasonable.

      If your system can run linux/bsd apps there's a beta client available than runs ~1.8x as fast as the standard one.

      http://einstein.phys.uwm.edu/app_test.php

      If you can run WINE, and get the 4.x boinc client working akos's C40 optimized app should give you a ~3x speed up.

      Otherwise until a new offcial app's released I think you're out of luck. IIRC there're only a few hundred solaris machines on the project total, so compiling beta apps for them isn't a priority. In the mean time, you might want to look for a project with more forgiving WU sizes or deadlines for your machine.

      Dotsch
      Dotsch
      Joined: 1 May 05
      Posts: 50
      Credit: 485389
      RAC: 396

      RE: If your system can run

      23 Apr 2006 23:07:20 UTC
      Message 29321 in response to message 29320
      (moderation:
        )

        Quote:
        If your system can run linux/bsd apps there's a beta client available than runs ~1.8x as fast as the standard one.


        No, a other OS is no option. Also it is a SPARC, so i386 binaries would not run on this system.

        Quote:
        If you can run WINE, and get the 4.x boinc client working akos's C40 optimized app should give you a ~3x speed up.


        So far I know, WINE only maps API calls. So it is not posible to let a i386 binary run on sparc.

        Quote:
        Otherwise until a new offcial app's released I think you're out of luck. IIRC there're only a few hundred solaris machines on the project total, so compiling beta apps for them isn't a priority. In the mean time, you might want to look for a project with more forgiving WU sizes or deadlines for your machine.


        I thought for einstein as primary or as backup project for this system. But this depends on the runtimes...

        bloed_brot
        bloed_brot
        Joined: 5 Apr 05
        Posts: 70
        Credit: 91124558
        RAC: 0

        My mate got hold of an old

        25 Apr 2006 9:14:39 UTC
        Message 29322
        (moderation:
          )

          My mate got hold of an old sun workstation. I will have a look at it on the week-end. Since I never had a sun workstation, I am very interested in seeing what this is all about. Also of course how well it copes with Einstein, hehe.

          I shall report back with some more information when at hand.

          :
          your thoughts - the ways :: the knowledge - your space
          :

          Bernd Machenschalk
          Bernd Machenschalk
          Moderator
          Administrator
          Joined: 15 Oct 04
          Posts: 4273
          Credit: 245277759
          RAC: 11859

          I have to admit that the

          25 Apr 2006 14:27:19 UTC
          Message 29323
          (moderation:
            )

            I have to admit that the current official Solaris App is indeed somewhat slow. I just wanted you to know that there is being worked on a new one, though it may take a while 'till it's out. It will probably be announced here.

            BM

            BM

            Dotsch
            Dotsch
            Joined: 1 May 05
            Posts: 50
            Credit: 485389
            RAC: 396

            RE: I have to admit that

            26 Apr 2006 15:19:48 UTC
            Message 29324 in response to message 29323
            (moderation:
              )

              Quote:
              I have to admit that the current official Solaris App is indeed somewhat slow. I just wanted you to know that there is being worked on a new one, though it may take a while 'till it's out. It will probably be announced here.


              Thank you very much, Bernd.
              Btw. is einstein floating, integer-, cache or memory(bandwith) intensive ?

              Kotulic Bunta
              Kotulic Bunta
              Joined: 22 Jan 05
              Posts: 33
              Credit: 859204
              RAC: 0

              I am happy to hear that a new

              27 Apr 2006 0:55:04 UTC
              Message 29325
              (moderation:
                )

                I am happy to hear that a new version is under development. It is really little bit slow - crunching times for 1 WU on my SPARC 1.0 GHz is around 106,000 seconds and on my SPARC 1.5GHz around 76,000 seconds.
                But better than nothing... :-)

                DanNeely
                DanNeely
                Joined: 4 Sep 05
                Posts: 1364
                Credit: 3562358667
                RAC: 109

                RE: I am happy to hear that

                27 Apr 2006 1:57:14 UTC
                Message 29326 in response to message 29325
                (moderation:
                  )

                  Quote:
                  I am happy to hear that a new version is under development. It is really little bit slow - crunching times for 1 WU on my SPARC 1.0 GHz is around 106,000 seconds and on my SPARC 1.5GHz around 76,000 seconds.
                  But better than nothing... :-)

                  That's painful. :( Between 2.5 and 3x slower than my athlon1500 running the stock client.

                  Bernd Machenschalk
                  Bernd Machenschalk
                  Moderator
                  Administrator
                  Joined: 15 Oct 04
                  Posts: 4273
                  Credit: 245277759
                  RAC: 11859

                  RE: Btw. is einstein

                  27 Apr 2006 14:56:25 UTC
                  Message 29327 in response to message 29324
                  (moderation:
                    )

                    Quote:
                    Btw. is einstein floating, integer-, cache or memory(bandwith) intensive ?


                    Mainly FPU. Once the FP code is fast enough, memory bandwith might come into play, but currently we're far from it.

                    BM

                    BM

                    DanNeely
                    DanNeely
                    Joined: 4 Sep 05
                    Posts: 1364
                    Credit: 3562358667
                    RAC: 109

                    RE: RE: Btw. is einstein

                    27 Apr 2006 22:10:35 UTC
                    Message 29328 in response to message 29327
                    (moderation:
                      )

                      Quote:
                      Quote:
                      Btw. is einstein floating, integer-, cache or memory(bandwith) intensive ?

                      Mainly FPU. Once the FP code is fast enough, memory bandwith might come into play, but currently we're far from it.

                      BM

                      While granting that it's heavily dependent on architecture, is "far from it" approximately a factor of a half dozen or so, a few dozen, hundreds, something even huger?

                      Comment viewing options

                      Select your preferred way to display the comments and click "Save settings" to activate your changes.

                      Forums › Cruncher's Corner