Announcement

Collapse
No announcement yet.

AMD EPYC Milan Performance Across 11 Different 2021 Linux Distributions

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

  • AMD EPYC Milan Performance Across 11 Different 2021 Linux Distributions

    Phoronix: AMD EPYC Milan Performance Across 11 Different 2021 Linux Distributions

    After looking recently at the FreeBSD 13.0 and DragonFlyBSD 6.0 performance on AMD EPYC 7003 "Milan" using a Tyan Transport CX GC68-B8036-LE server, the next round of benchmarking from this server with AMD EPYC 7543 32-core processor was looking at its support (all tested 2021 Linux distributions were running fine on this latest-generation AMD server) and performance across 11 current Linux distribution releases from Arch, CentOS, Clear Linux, Debian, Fedora, Rocky Linux, Ubuntu, and openSUSE.

    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
    Is it just me, or lately it seems Ubuntu is always the slowest in these benchmarks? Really wonder why other distributions consistently outperform it.

    Comment


    • #3
      It's nice to see INTEL 100% behind AMD EPYC with ClearLinux :-)

      Comment


      • #4
        Originally posted by user1 View Post
        Is it just me, or lately it seems Ubuntu is always the slowest in these benchmarks? Really wonder why other distributions consistently outperform it.
        It's bloated, and not much attention to performance. They are more focused on UI doodads and functionality. Where Clear Linux is really only focusing on benchmark performance.

        Comment


        • #5
          Originally posted by phoronix
          When running all of the Intel oneAPI benchmarks available via the Phoronix Test Suite there wasn't too much difference besides Ubuntu 10.10 being much slower than the other ten Linux distributions.
          No doubt it is, but I suspect the pattern was too strong, and this was meant to say 'Debian 10.10'

          Comment


          • #6
            Originally posted by user1 View Post
            Is it just me, or lately it seems Ubuntu is always the slowest in these benchmarks? Really wonder why other distributions consistently outperform it.
            Linux distributions performance is primarily to these two things:
            • Compiler and compilation flags
            • Kernel configuration and sysctls - you can easily overoptimize the latter to the detriment in performance
            And it has very little to do with "bloat" and "UI doodads" mentioned earlier. You can have a ton of things installed by default but if they are not running or sleeping, they will unlikely to affect performance in any meaningful way. The test system contains a boatload of RAM.

            Comment


            • #7
              Originally posted by cb88 View Post

              It's bloated, and not much attention to performance. They are more focused on UI doodads and functionality. Where Clear Linux is really only focusing on benchmark performance.
              That's not the case. Ubuntu usually uses the slowest settings like Powersave governor.

              Comment


              • #8
                Originally posted by Volta View Post

                That's not the case. Ubuntu usually uses the slowest settings like Powersave governor.
                That is the case... and is kind of the whole point.

                Comment


                • #9
                  Um, the comparison is missing Gentoo...
                  SCNR.
                  *escapes laughing*

                  But interesting anyway, so people can check for common workloads they might have similar to a benchmark and then choose distribution.
                  Stop TCPA, stupid software patents and corrupt politicians!

                  Comment


                  • #10
                    Originally posted by cb88 View Post

                    That is the case... and is kind of the whole point.
                    It has nothing common with bloat. They often make stupid decisions. Previously they chose IO scheduler not suited for desktops.

                    Comment

                    Working...
                    X