Last week when delivering some CachyOS benchmarks against Fedora 43 and Ubuntu 25.10 on the Framework Desktop with AMD Ryzen AI Max+, a few Phoronix readers wrote in with the question or belief that openSUSE Tumbleweed would better perform against CachyOS given the distribution’s select x86_64-v3 packages and other advantages. As it’s been a while since running any benchmarks of the rolling-release openSUSE Tumbleweed, here are those benchmarks now in the mix for seeing how the performance compares.

  • Björn@swg-empire.de
    link
    fedilink
    arrow-up
    16
    ·
    14 hours ago

    Long story short on a geo mean basis for all the tests that ran successfully on each of the tested Linux distributions, openSUSE Tumbleweed was slightly ahead of Fedora Workstation 43 for a second place finish but CachyOS still delivering notably better performance overall.

    My takeaway was that Tumbleweed blew Cachy out of the water in Super Tux Kart Vulkan, the only important benchmark.

    • brucethemoose@lemmy.world
      link
      fedilink
      arrow-up
      5
      ·
      edit-2
      14 hours ago

      Yeah.

      Its so disparate (and so different from all the OpenGL results) it must be some config thing. What Vulkan switch is openSUSE flipping that all other other distros missed?

      • Björn@swg-empire.de
        link
        fedilink
        arrow-up
        3
        ·
        14 hours ago

        I’d also like to know what leads to Ubuntu doing so well in the PHP and Python benchmarks.

        • brucethemoose@lemmy.world
          link
          fedilink
          arrow-up
          4
          ·
          edit-2
          13 hours ago

          That could be versioning?

          For example, Arch Linux (CachyOS) sticks to one Python version systemwide, and doesn’t upgrade it until the whole Arch repo is ready to transition, while in Ubuntu many Python version exist in parallel. The Phoronix bench presumably uses the latest one. Perhaps Python 3.14 (and associated libraries) have some dramatic performance gains in certain benchmarks?


          I know there are some Python ‘hacks’ as well. Clear Linux, for instance, used to ship Python with some custom build flags and code patches that forced AVX2 and some other stuff, but I think Cachy pulled these in?