Announcement

Collapse
No announcement yet.

Linux 4.10 To Linux 4.15 Kernel Benchmarks

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

  • Linux 4.10 To Linux 4.15 Kernel Benchmarks

    Phoronix: Linux 4.10 To Linux 4.15 Kernel Benchmarks

    The ThinkPad X1 Carbon has been enjoying its time on Linux 4.15. In addition to the recent boot time tests and kernel power comparison, here are some raw performance benchmarks looking at the speed from Linux 4.10 through Linux 4.15 Git...

    http://www.phoronix.com/scan.php?pag...5-Kernel-Tests

  • #2
    Thanks a lot for this Michael!

    Some food for thought: personally, I think it'd be more interesting to have some LTS kernels instead of the EOL ones.

    What you have shown here is of course very interesting in showing the progress in a chronologically consistent manner. But, for instance, I'd find more value in seeing how 4.9 (LTS) is doing instead of 4.10, which has been EOLed for 6 months now.

    Anyhow, just putting an idea out there. Keep it up!

    Comment


    • #3
      That 4.12 kernel was flying through all of the I/O tests! Someone needs to figure out what was different about the I/O in the 4.12 kernel and reapply the changes to the latest kernels!

      Comment


      • #4
        One can certainly say: WEIRD

        Comment


        • #5
          How can the IO throughput in this article be the best (highest) for kernel version 4.12, when boot time is worst (longest) for kernel version 4.12 based on the article below?
          https://www.phoronix.com/scan.php?pa...rnel-Boot-Time

          Comment


          • #6
            Oh wow, when I saw the I/O benchmarks on 4.12 I immediately wondered if it might solve the bogging down issues during large file transfers with the other 4.10 and later kernels. It did help a bit, in fact I'm actually typing this while transferring a 130GB file which I can't do with any other kernel, no matter the IO tuning. I don't see any transfer speed improvement though.

            Comment


            • #7
              These types of reviews are exactly why I'm happy to be a premium member. It's basic stuff, but no other sites ever do these types of comparisons. It's always "the latest version".

              Comment


              • #8
                Originally posted by AndyChow View Post
                These types of reviews are exactly why I'm happy to be a premium member. It's basic stuff, but no other sites ever do these types of comparisons. It's always "the latest version".
                Exactly AndyChow. I honestly don't know how Micheal does it. Every day there's useful and interesting information, and it often helps me with immediate issues or gives me better insight into future ones.

                Comment


                • #9
                  Originally posted by Herem View Post
                  That 4.12 kernel was flying through all of the I/O tests! Someone needs to figure out what was different about the I/O in the 4.12 kernel and reapply the changes to the latest kernels!
                  I know 4.13 had many regressions performance wise in blk_mq, but they should be fixed in 4.15 and Michael is using CFQ so not sure...

                  Comment


                  • #10
                    After noticing a significant difference in results for the himeno test compared to a year or more ago, and narrowing it down to between kernel 4.12 and 4.13, I searched and found this very relevant article. Thanks.

                    Comment

                    Working...
                    X