Announcement

Collapse
No announcement yet.

Liquorix Kernel Benchmarked Against Linux 4.2, Linux 4.3 Kernels

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

  • Liquorix Kernel Benchmarked Against Linux 4.2, Linux 4.3 Kernels

    Phoronix: Liquorix Kernel Benchmarked Against Linux 4.2, Linux 4.3 Kernels

    Recently there were a number of requests about testing the latest state of Liquorix, the self-prcolaimed "better distro kernel" that is an optimized version of the Linux kernel with extra patches that makes it optimal for desktop, multimedia, and gaming workloads. Here's some fresh Liquorix vs. mainline Linux kernel performance benchmarks...

    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
    I think that when you are comparing two different types of kernel, you shouldn't be using standard benchmarks to measure performance but only test for differences in latency, as that's primarily what a desktop-oriented kernel is going to do: improve latency.

    Comment


    • #3
      I partially disagree with you nmstick.
      I think to be relevant you'd need both, because you'd want to know how much performance you're trading for a lower latency.
      (Hopefully none, but trading too much would be bad as well)

      Comment


      • #4
        Basically 1000 Hz kernels need more for themself. That'tat u woukd try to get from your wolk!?

        Comment


        • #5
          I may be wrong, but what I understand about most of those kernel patches that improve Desktop and Gaming experience, what they do is to improve the overall responsiveness of desktop session when system is doing other things. If running a single "Job" may or may not show any improvement at all. And that's exactly what your test shows.

          But for example if you do a copy operation in the background, while your gaming session (or music session) is at full swing, you will notice no latency or shattering of sound, more responsive mouse and overall the feel of a more responsive system.

          But more frames... no no no. Maybe a minimal increase/decrease but that's not what those patches are for.
          Last edited by darkcoder; 13 October 2015, 10:06 PM.

          Comment


          • #6
            Originally posted by darkcoder View Post
            I may be wrong, but what I understand about most of those kernel patches that improve Desktop and Gaming experience, what they do is to improve the overall responsiveness of desktop session when system is doing other things. If running a single "Job" may or may not show any improvement at all. And that's exactly what your test shows.

            But for example if you do a copy operation in the background, while your gaming session (or music session) is at full swing, you will notice no latency or shattering of sound, more responsive mouse and overall the feel of a more responsive system.

            But more frames... no no no. Maybe a minimal increase/decrease but that's not what those patches are for.
            I wonder how one would benchmark this.. Would it be possible to measure the time between input and when the cursor is repainted, and logging those results while running different loads? Sounds like it would be hard to get the timestamps correct since I assume the two parts are in different software stacks..

            Comment


            • #7
              Originally posted by nilssab View Post

              I wonder how one would benchmark this.. Would it be possible to measure the time between input and when the cursor is repainted, and logging those results while running different loads? Sounds like it would be hard to get the timestamps correct since I assume the two parts are in different software stacks..
              This is called measuring motion to photon latency (MTP) and is regularly measured in VR prototyping.

              Comment


              • #8
                Originally posted by nilssab View Post

                I wonder how one would benchmark this.. Would it be possible to measure the time between input and when the cursor is repainted, and logging those results while running different loads? Sounds like it would be hard to get the timestamps correct since I assume the two parts are in different software stacks..

                Comment


                • #9
                  but ... I build kernels [+ its infrastructure - delivered by our NeteXt'73], which provides both high responsiveness, productivity and energy efficiency. Examples below:


                  our site: http://www.netext73.pl

                  Performance:

                  http://www.netext73.pl/2015/08/cs-go...17-poleng.html


                  Energy efficiency:

                  https://dl.dropboxusercontent.com/u/...roadwell-1.png

                  https://dl.dropboxusercontent.com/u/...roadwell-2.png

                  https://dl.dropboxusercontent.com/u/...roadwell-3.png

                  https://dl.dropboxusercontent.com/u/...roadwell-4.png

                  https://dl.dropboxusercontent.com/u/...roadwell-5.png

                  https://dl.dropboxusercontent.com/u/...ny/v42.5-1.png


                  Responsivnes:

                  https://dl.dropboxusercontent.com/u/...0.9-resp-1.png

                  https://dl.dropboxusercontent.com/u/...0.9-resp-2.png

                  https://dl.dropboxusercontent.com/u/...0.9-resp-3.png

                  https://dl.dropboxusercontent.com/u/...0.9-resp-4.png

                  ... old ones ... now is much better

                  Moje rozwiązania: Kernel 3.13.0-rc7-ext73-f1-28.0-brazos-ags-cfs + APM 3.2 + wpisy w Grub oraz Linux-firmware/amd microcode. Tu prezentacja działania kernela...

                  Last edited by ext73; 14 October 2015, 08:56 AM.

                  Comment


                  • #10
                    How am I supposed to parse the table in the article? What do the numbers mean? Sometimes the highest values are red and sometimes they are green. I assume that the values are only relevant to the individual test, but some guidance would be nice. A link to where the tests are described would be good to include in the article.

                    Comment

                    Working...
                    X