Announcement

Collapse
No announcement yet.

Linux 4.1 Offers Potentially Dazzling Performance

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

  • Linux 4.1 Offers Potentially Dazzling Performance

    Phoronix: Linux 4.1 Offers Potentially Dazzling Performance

    Besides presenting a lot of new kernel features and functionality, the upcoming Linux 4.1 kernel release is potentially very exciting if you're an owner of certain classes of Intel hardware that offer better performance under this new kernel -- and in some cases, better battery life. Here's some tests from yet another system I found exhibiting some promising results from this new 2015 summer kernel version.

    http://www.phoronix.com/vr.php?view=21746

  • #2
    Do you Micheal have any idea why i have better results in Xonotic phoronix-test-suite if i have network off?

    Those minimal fps became higher something like +5 or 10%, does xonotic or suite check something while benchmarking?

    I know for about audio setup overhead, but did't expect network does this too

    Comment


    • #3
      Half kernel build time is crazy!

      Comment


      • #4
        Originally posted by doom_Oo7 View Post
        Half kernel build time is crazy!
        It probably used half amount of cores previously

        edit: For about xonotic via suite... yep suite does something with network - xonotic alone benhmark does not do that
        Last edited by dungeon; 06-06-2015, 04:17 PM.

        Comment


        • #5
          Originally posted by dungeon View Post
          Do you Micheal have any idea why i have better results in Xonotic phoronix-test-suite if i have network off?

          Those minimal fps became higher something like +5 or 10%, does xonotic or suite check something while benchmarking?

          I know for about audio setup overhead, but did't expect network does this too
          Only thing I can think of would be Xonotic checking fo network connectivity for updares of Xonotic... but if it really takes ~5 FPS, then wow,.
          Michael Larabel
          http://www.michaellarabel.com/

          Comment


          • #6
            He, he, 5 fps here, 5 fps there... top performance

            Yeah it check in the middle of the benchmark if version is up to date, that is a bug anyway because it affects result... there might be some cvar to turn that off.

            Comment


            • #7
              Michael,
              one thing that I find missing in these charts is a summary score.
              I think it would be nice to be able to get a summary score to see how it went overall.
              You could make the last stable kernel at 100% and then compute each test score and average them at the end (that's just one way of course)

              Comment


              • #8
                Originally posted by doom_Oo7 View Post
                Half kernel build time is crazy!
                The word I'd use is "unbelievable".

                Comment


                • #9
                  Originally posted by geearf View Post
                  Michael,
                  one thing that I find missing in these charts is a summary score.
                  I think it would be nice to be able to get a summary score to see how it went overall.
                  You could make the last stable kernel at 100% and then compute each test score and average them at the end (that's just one way of course)
                  From the OpenBenchmarking.org result file you can look at the geometric sum and other options via the analyze tab... For any other options, patches are welcome as I'm no stats expert, etc, while the software is open-source.
                  Michael Larabel
                  http://www.michaellarabel.com/

                  Comment


                  • #10
                    Originally posted by geearf View Post
                    Michael,
                    one thing that I find missing in these charts is a summary score.
                    I think it would be nice to be able to get a summary score to see how it went overall.
                    You could make the last stable kernel at 100% and then compute each test score and average them at the end (that's just one way of course)

                    Yes, I've been suggesting this for long

                    MIchael added a lot of nice functionality, like normalization of results. You can play on the Analyze tab:
                    http://openbenchmarking.org/result/1...hgv=Linux+3.18

                    However, the resulting geometric means make no sense. There is some bug. If I take, for instance, all the numbers in the last column, for these tests, the geometric average I get is arround 1.14, which means that kernel 4.1 is around 14% faster than 3.18 according to these tests. However, he's getting 0.96, that is, 4% slower, it doesn't make any sense. I wish I had time to debug and fix this. Sigh.

                    But back to your point, yes, if I were you, I would show a summary of that sort, besides some important plots.

                    Comment

                    Working...
                    X