Announcement

Collapse
No announcement yet.

Windows Server 2019 vs. Linux vs. FreeBSD Performance On A 2P EPYC Server

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

  • Windows Server 2019 vs. Linux vs. FreeBSD Performance On A 2P EPYC Server

    Phoronix: Windows Server 2019 vs. Linux vs. FreeBSD Performance On A 2P EPYC Server

    When Microsoft rolled out their Windows 10 October 2018 Update they also released Windows Server 2019. Now over the slower holiday period I am finally getting caught up in benchmarking Windows Server 2019. For this initial benchmark comparison is a look at the Microsoft Windows Server 2019 performance against a handful of Linux distributions as well as FreeBSD 12.0 for seeing how this latest Windows Server performance compares on a dual AMD EPYC 7601 server.

    http://www.phoronix.com/vr.php?view=27311

  • #2
    Strange how Clear fared so poorly this time around. I wonder if there's some issue with there being multiple CPUs, where it's trying to cross-communicate between them, resulting in terrible performance.

    Comment


    • #3
      Michael, you own the 64-core 128-thread 512GB-ram 9TB-disk machine or is it a 3rd party machine?

      Comment


      • #4
        Originally posted by atomsymbol View Post
        Michael, you own the 64-core 128-thread 512GB-ram 9TB-disk machine or is it a 3rd party machine?
        AMD sent it over
        Michael Larabel
        http://www.michaellarabel.com/

        Comment


        • #5
          Originally posted by schmidtbag View Post
          Strange how Clear fared so poorly this time around. I wonder if there's some issue with there being multiple CPUs, where it's trying to cross-communicate between them, resulting in terrible performance.
          It mostly faired poorly on I/O but may also be due to some issues on AMD hardware but unsure. I don't believe it's a multi-socket issue as on the Intel Xeon it performs great albeit different storage. Anyhow, will have the Xeon tests up in a few days.
          Michael Larabel
          http://www.michaellarabel.com/

          Comment


          • #6
            Typo:

            Originally posted by phoronix View Post
            in fact winning in the "Solitare" benchmark.

            Comment


            • #7
              Clear linux should not be so racist /joking. Intel didn't plan past 28 cores at 5Ghz. The acoladies are in delivering a girl her CPU that works.
              Last edited by rukur; 12-24-2018, 12:47 PM.

              Comment


              • #8
                I noticed some benchmarks weren't run on all OSes - for instance, CentOS is missing from most of the benchmarks on page 7. Was this due to build difficulties?

                Comment


                • #9
                  Originally posted by jbrnd View Post
                  I noticed some benchmarks weren't run on all OSes - for instance, CentOS is missing from most of the benchmarks on page 7. Was this due to build difficulties?
                  Yes like e.g. OSBench requires Meson+Ninja for its build system but that not being available on CentOS 7. Or other issues with the libraries particularly on CentOS/EL7 due to being much older than what is needed by some tests.
                  Michael Larabel
                  http://www.michaellarabel.com/

                  Comment


                  • #10
                    Typically install hangs (Windows Server xxxx) are caused by device/bus enumeration failures. Depending on the BIOS supplier and the board maker, and the types of options installed, the BIOS may present the wrong bus or device first to the OS during setup causing a hang.

                    As an example, we see pre-installed Emulex adapters in BUS3/SLOT1 on certain HP models fail to install because the BIOS enumerates BUS 3 first and Windows gets lost. Moving the Emulex to a secondary bus for enumeration fixes it.

                    I am not saying that is the issue here, I am just saying this is what I have seen that blocks a Windows Server install from finishing.

                    Comment

                    Working...
                    X