Ubuntu 18.04 Versus Six Other Linux Distributions On AMD EPYC

Written by Michael Larabel in Operating Systems on 15 March 2018 at 07:00 AM EDT. Page 1 of 4. 12 Comments.

With Ubuntu 18.04 LTS set to be released next month and its final package configuration quickly falling into place, we have begun firing up some benchmarks for seeing how this Ubuntu 18.04 "Bionic Beaver" release is comparing to various other Linux distributions. Up first as part of this series of benchmarks is using an AMD EPYC workstation/server for seeing how the Ubuntu 18.04 LTS performance compares to six other Linux distributions.

This latest Linux distribution benchmarking roundabout on Phoronix was done with an AMD EPYC 7551 32-core / 64-thread processor with the GIGABYTE MZ31-AR0 motherboard, 8 x 4GB DDR4-2666 ECC Registered Memory, NVIDIA GeForce GTX 1070 Ti (albeit not used as part of today's test selection), ASpeed display hardware from the Gigabyte EPYC server motherboard, and 500GB Samsung 960 EVO NVMe SSD.

The seven Linux distributions tested on this Gigabyte + AMD EPYC system included in alphabetical order: Clear Linux, Debian Testing, Fedora Workstation, Manjaro, Ubuntu 16.04 LTS, Ubuntu 18.04 LTS, and openSUSE Tumbleweed.

Clear Linux 21260 - While Intel's distribution out of their Open-Source Technology Center is designed for Intel hardware, we have found it to run mighty fast and well on modern AMD x86_64 hardware too, thus including it for this EPYC testing. This rolling-release distribution as of build 21260 was using the Linux 4.15.7 kernel, GCC 7.3.0 compiler, defaulted to the Kyber I/O scheduler on this NVMe drive (all other Linux distributions tested defaulted to no I/O scheduler with this NVMe storage), and defaults to the ACPI CPUFreq driver with its performance governor.

Debian Testing - Debian 9 stable couldn't be tested as once installed it was repeatedly hanging while trying to install the benchmarks. Debian 10 Buster testing went fine on this EPYC system. Debian Testing currently has the Linux 4.14 kernel and GCC 7.3.0. The kernel provided full AMD Retpoline support but no user pointer sanitization or OSB support for Spectre V1 mitigation.

Fedora Workstation 27 - Fedora 27 was running fine on this box but CentOS 7 meanwhile couldn't be tested as once installed its EFI configuration was failing to work on this system. Fedora 27 with updates has the Linux 4.15 kernel, and GCC 7.3.1.

Manjaro 17.1.6 - Manjaro was used this time over Antergos due to that Arch derivative having display issues with the ASpeed hardware. Manjaro 17.1.6 with updates is on the Linux 4.14 kernel with GCC 7.3.0.

Ubuntu 16.04 LTS - This current Ubuntu Long Term Support release is at the Linux 4.13 kernel with its hardware enablement stack while using the GCC 5.4.0 compiler. Ubuntu 16.04's updated kernel has full AMD Retpoline support but is using the Observable Speculation Barrier for Spectre V1 mitigation rather than the user pointer sanitization found normally on the other tested distributions (sans no V1 coverage with Debian Testing).

Ubuntu 18.04 LTS - Ubuntu 18.04 today and for its final release is making use of the Linux 4.15 kernel and GCC .7.3.

OpenSUSE Tumbleweed - The rolling-release of openSUSE was also tested and was the only distribution this time around not using EXT4 but continues defaulting to XFS for home and Btrfs for the root file-system. OpenSUSE Tumbleweed is currently making use of the Linux 4.15 kernel and GCC 7.3.1 compiler.

Ubuntu 18.04 vs. Other Linux Distros On AMD EPYC

On these seven Linux distributions they were cleanly installed each time in an "out of the box" manner to represent what any new user would find for the default OS configuration as set by the vendor. Via the Phoronix Test Suite a range of server, workstation, and general CPU/system benchmarks were run given the AMD EPYC 7551 platform being used today. In our other Ubuntu 18.04 comparisons forthcoming on desktop hardware will be the more gaming-type benchmarks.


Related Articles