Announcement

Collapse
No announcement yet.

Clear Linux On The OnLogic Karbon 700 Boosted Performance By 13% Over Ubuntu With 141 Benchmarks

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

  • Clear Linux On The OnLogic Karbon 700 Boosted Performance By 13% Over Ubuntu With 141 Benchmarks

    Phoronix: Clear Linux On The OnLogic Karbon 700 Boosted Performance By 13% Over Ubuntu With 141 Benchmarks

    Last month we reviewed the OnLogic Karbon 700 as a passively-cooled, industrial-grade PC powered by an eight-core / sixteen-thread Intel Xeon, 16GB of RAM, 512GB NVMe storage, and a plethora of connectivity options in suiting to industrial use-cases. The performance was great and even the thermal performance was very good for being a fan-less PC. In seeing how well other Linux distributions were panning out on the Karbon 700, I tested five popular Linux distributions on the Xeon Coffee Lake system and once again Intel's performance-optimized Clear Linux squeezed out much more performance potential.

    http://www.phoronix.com/vr.php?view=28614

  • #2
    Distros need to take performance serious.
    For the SQLite results a simple pass of:

    Code:
    export CFLAGS+=" -O3 -falign-functions=32 -ffat-lto-objects -flto=4 -fno-math-errno -fno-semantic-interposition -fno-trapping-math
    export CXXFLAGS+=" $CFLAGS"
    export LDFLAGS+=" $CFLAGS"
    Will improve it's performance.

    For Glibc for example ClearLinux applied patches 2 years ago to beat the competition, but today most of those patches are already upstream (by contribution of Clearlinux) since version 2.28 patches still left are:

    ldconfig-format-new.patch #Make ldconfig default to "new" cache format
    malloc_tune.patch # tune malloc a bit more
    use_madv_free.patch # tune for madvise free usage
    fix-ld-audit-performance.patch #Address performance issues with LD_AUDIT (rtld-audit)
    nostackshrink.patch # improve clone performance
    mathlto.patch #LTO math functions and pass CFLAGS-.o += -fno-stack-protector -falign-functions=32 to it.
    spinaphore.patch # Revise spin rounds in pthread mutex lock
    alternate_trim.patch # alternate trim for performance (malloc code)
    spin-smarter.patch # spin smarter on pthread_mutex_lock.c

    With a pass to the compiler of:
    Code:
    export CFLAGS+=" -O3 -march=westmere -mtune=skylake -fPIC -falign-functions=32 -mbranches-within-32B-boundaries"
    export CXXFLAGS+=" $CFLAGS"
    If you don't have these architectures then just pass -march=native without -mtune

    Visit https://github.com/clearlinux-pkgs for more.
    Last edited by r08z; 12-24-2019, 07:23 PM.

    Comment


    • #3
      Speed may not be the only parameter to consider, Intel taught it with its CPUs. There is a far more important aspect, that of security.
      However on my PC I have not noticed substantial differences in terms of speed between Tumbleweed (which I usually use) and Clear Linux, indeed Tumbleweed seems faster, but I think it depends on the fact that the applications of Clear Linux are in flatpak and snap, which they are much slower. So all this optimization goes to hell for desktop use!

      Comment

      Working...
      X