Announcement

Collapse
No announcement yet.

Dirndl On Intel's Sandy (Bridge) Is Hot

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

  • Dirndl On Intel's Sandy (Bridge) Is Hot

    Phoronix: Dirndl On Intel's Sandy (Bridge) Is Hot

    To some frustration, the big software announcement that's codenamed Dirndl hasn't yet hit wire. I'm told the company is deciding within the next couple of hours whether they want to issue a press release on a Friday or hold off until next week, but regardless of the day of week it will generate a lot of attention due to this game-changing move. Soon as I hear it's hit the wire, a four-page Phoronix article is in the queue...

    http://www.phoronix.com/vr.php?view=OTU1Mg

  • #2
    Sandy's quite good in a Dirndl, as expected. Now to wait a few hours and can hopefully then blow the top on this story...
    Haha, just looked up Dirndl on google images. Makes sense now

    Comment


    • #3
      ENZO

      here is hoping ENZO is also part of the deal...

      Comment


      • #4
        I hope that dress can optimize radeon opensource drivers.
        if their performance is boosted at least to 80-75% of proprietary crap I will at last get decent amd card. At same time, you hope amd itself should have done that...

        Comment


        • #5
          It would seem obvious to me that what is hiding behind this codename is either a new compiler, or a new distro, but the latter would seem unlikely since the catchphrase here seems to be "open-source". Also, the tests are mostly computational in nature. Also i thought it could be a new kernel, but i doubt this the most.

          Comment


          • #6
            Now that I know what "Dirndl" means I have one more reason to go to Oktoberfest


            Back on topic, I would like to see some benchmarks on crappy hardware like mine too (i.e. single core 5yo CPU). Not only with this "Dirndl" but more in general.
            I shouldn't be too difficult to get nowadays. Just go to some recycling center and pick up the first computer you see.

            Comment


            • #7
              I noticed that except for the NAS benchmark you are using the same small set of benchmarks as you did for the other two articles.

              Now the skeptic in me awakes and says these benchmarks are cherry picked. Either because there are problems getting 'Dirndl' to run the other benchmarks, or just to make 'Dirndl' look good, i.e., the benchmark results of 'Dirndl' aren't that impressive on the other benchmarks.

              Prove me wrong and post some x264 benchmarks!
              Last edited by monraaf; 06-10-2011, 11:11 AM.

              Comment


              • #8
                Originally posted by monraaf View Post
                Prove me wrong and post some x264 benchmarks!
                Well, x264 benchmarks would be pointless since Phoronix doesn't disable the hand-optimized assembly when benchmarking it. Practically all performance hotspots in x264 has specific hand optimized assembly written for it and as such will not be optimized in any way by a compiler.

                As for the cherry-picking, I'm with you there. Still, back when it was just the himeno pressure test I was taking the performance with a grain of salt (given that there's been some very huge differences in that particular test) but now we've also seen C-Ray, TSCP Chess Ai show great performance so atleast we have a slightly larger testbed now indicating that there is alot of performance increase to be had for certain types of applications.

                However, we don't know what compiler options these NEW tests where done with, chances are they are unfair to the standard compiler flags used in the Phoronix benchmarks (on the other hand, maybe they're exactly the same), we just don't have the data as of now.

                As for what this release pertains, speedups like 2x, 2.5x etc indicates to me that this is a combined cpu+gpu (gpgpu) compiler. I find it hard to believe that ekopath would deliver such a large performance increase over gcc,clang/llvm. Granted, I've never used ekopath but I've used icc and I've never had anything like 2x the performance with it.

                Comment


                • #9
                  Is Dirndl the official codename of this thing or is it something Michael made up?

                  Comment


                  • #10
                    Originally posted by wea0 View Post
                    Is Dirndl the official codename of this thing or is it something Michael made up?
                    No idea. These articles are absurd.

                    Comment

                    Working...
                    X