Announcement

Collapse
No announcement yet.

HPL-GPU 1.0 Released For GPU Super Computing

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

  • HPL-GPU 1.0 Released For GPU Super Computing

    Phoronix: HPL-GPU 1.0 Released For GPU Super Computing

    Matthias Bach, a researcher and developer on high-performance computing with the Goethe Universitaet Frankfurt am Main, has written in to inform us of this German university's release of HPL-GPU 1.0. The HPL-GPU software package is a rewritten version of the LINPACK library that is re-engineered to run atop CALDGEMM, which is a DGEMM implementation developed at this university designed to run atop the latest AMD graphics processors...

    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

    Matthias Bach also mentioned to us in his email, "The really interesting thing however is, that this code can run AMD GPUs (Cypress Type, we use AMD 5870) at 100% load for hours, a load that except of our Linpack/DGEMM only Furmark can cause...The code also stomps NVIDIA, reaching 497 flops DGEMM on the HD5870 where NVIDIA only reaches around 300 on the much more expensive Tesla systems."
    Not surprising considering

    The code is also engineered for the Radeon HD 5000 "Evergreen" series of graphics cards.
    Too bad he didn't mention which Tesla card he was using for comparison.

    Comment


    • #3
      i wonder why they didnt add support for the open driver :O

      Comment


      • #4
        Its not mature enough. It may never be for this type of load. As a matter of fact it is exactly this type of load that the catalyst driver was designed for.

        I hate to admit it, but I was once a firm believer in an open source only strategy. I said it here on this forum many times. But in light of this, I admit that a closed source optimized binary is better overall in this situation.

        I for one am glad the binary exists in this particular case, because we woulnt have this new open source code without it.

        Comment


        • #5
          Originally posted by duby229 View Post
          Its not mature enough. It may never be for this type of load. As a matter of fact it is exactly this type of load that the catalyst driver was designed for.

          I hate to admit it, but I was once a firm believer in an open source only strategy. I said it here on this forum many times. But in light of this, I admit that a closed source optimized binary is better overall in this situation.

          I for one am glad the binary exists in this particular case, because we woulnt have this new open source code without it.
          EDIT: Just to clarify, I still believe the open source driver can surpass the closed driver. I still think it would be sweet justice if one day the open driver could perform exactly this type of load.

          But for now it cant, so closed driver FTW.

          Comment


          • #6
            Originally posted by duby229 View Post
            Its not mature enough. It may never be for this type of load. As a matter of fact it is exactly this type of load that the catalyst driver was designed for.
            Yep.

            I hate to admit it, but I was once a firm believer in an open source only strategy. I said it here on this forum many times. But in light of this, I admit that a closed source optimized binary is better overall in this situation.

            I for one am glad the binary exists in this particular case, because we woulnt have this new open source code without it.
            Heh... There is little reason that the FOSS driver won't eventually catch up. Surpass might be a bit of a stretch, but there's little reason to believe that we won't get there. They said the same thing about our OS choice back in earlier days, remember...

            Comment


            • #7
              Originally posted by Svartalf View Post
              Yep.



              Heh... There is little reason that the FOSS driver won't eventually catch up.
              Actually, I can think of four interconnected reasons:
              1. Lack of time
              2. Lack of manpower
              3. Lack of money
              4. Design for ease of implementation rather than performance

              Surpass might be a bit of a stretch, but there's little reason to believe that we won't get there. They said the same thing about our OS choice back in earlier days, remember...
              When did we surpass Windows in usage? Or do you mean something else?

              Comment


              • #8
                Originally posted by BlackStar View Post
                Actually, I can think of four interconnected reasons:
                1. Lack of time
                2. Lack of manpower
                3. Lack of money
                4. Design for ease of implementation rather than performance



                When did we surpass Windows in usage? Or do you mean something else?
                He's talking about performance. But those days are before my time. I didnt come to Linux until the beginning of the 2.6 cycle. So it's pretty much always been stable and fast from my perspective.

                Comment


                • #9
                  Originally posted by duby229 View Post
                  He's talking about performance. But those days are before my time. I didnt come to Linux until the beginning of the 2.6 cycle. So it's pretty much always been stable and fast from my perspective.
                  lol

                  I run the latest kde with the latest open source AMD/ati video card, so my linux system is neither stable nor fast...

                  Comment


                  • #10
                    Originally posted by d4ddi0 View Post
                    lol

                    I run the latest kde with the latest open source AMD/ati video card, so my linux system is neither stable nor fast...
                    but that's actually your own fault. Running latest git tree won't always give you best performance and least problems. XD

                    Comment

                    Working...
                    X