Announcement

Collapse
No announcement yet.

Ubuntu 22.10 Bringing Some Performance Uplift For Intel Xeon Scalable

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

  • Ubuntu 22.10 Bringing Some Performance Uplift For Intel Xeon Scalable

    Phoronix: Ubuntu 22.10 Bringing Some Performance Uplift For Intel Xeon Scalable

    Now that Ubuntu 22.10 is into its feature freeze and its Linux 5.19 based kernel landed as well as moving to the GCC 12.2 compiler, I've begun testing this forthcoming Ubuntu Linux (non-LTS) release on more systems. For the current-flagship Intel Xeon Platinum 8380 "Ice Lake" processors, Ubuntu 22.10 does deliver some performance advantages over the current Ubuntu 22.04.1 LTS release. However, Ubuntu 22.10 still trails in a distance behind Intel's own Clear Linux platform for the most aggressive out-of-the-box performance.

    https://www.phoronix.com/review/ubun...-early-icelake

  • #2
    Charts have error, they say 20.04.1 instead of 22.04.1

    Comment


    • #3
      Holy shit, Clear Linux is killing it!

      Comment


      • #4
        Michael

        Typo on page 3

        "was competing ell with" should be "was competing well with".

        Comment


        • #5
          Originally posted by wooque View Post
          Charts have error, they say 20.04.1 instead of 22.04.1
          Whoops yes, thanks.
          Michael Larabel
          https://www.michaellarabel.com/

          Comment


          • #6
            Originally posted by Anux View Post
            Holy shit, Clear Linux is killing it!
            It's an only distro here using performance governor.

            Comment


            • #7
              Originally posted by Volta View Post

              It's an only distro here using performance governor.
              Yeah I forgot.

              Comment


              • #8
                > Much of the lead with Clear Linux is due to opting for the "performance" CPU frequency scaling governor by default, the tuned CFLAGS/CXXFLAGS by default in the name of better performance, and also a ton of smaller individual optimizations to different packages from tuning to new patches.

                I'm not sure how true this still is. Obviously, not using a "bad" governor can be extremely significant in many workloads; but aside from that the bulk of Clear's performance seems to be from CPU-specific library replacements (e.g. using SSE/AVX instructions for mempcy, memset, etc) rather than gcc flags.

                Comment

                Working...
                X