Intel Continues To Show AMD The Importance Of Software Optimizations: 16% More Ryzen 9 9950X Performance

Written by Michael Larabel in Operating Systems on 20 August 2024 at 10:30 AM EDT. Page 2 of 7. 42 Comments.
miniBUDE benchmark with settings of Implementation: OpenMP, Input Deck: BM1. CachyOS was the fastest.
miniBUDE benchmark with settings of Implementation: OpenMP, Input Deck: BM1. CachyOS was the fastest.

Right from the start there was an interesting observation: with the miniBUDE HPC app there was clearly better performance for all of the Linux distributions besides Ubuntu 24.04 LTS. Even Ubuntu 24.10 was offering a nice bump to performance inline with the other Linux distributions. What's going on here? All of the Linux distributions besides Ubuntu 24.04 LTS are using the GCC 14 compiler while Ubuntu 24.04 LTS relies on GCC 13. GCC 14 has the Znver5 tuning and various other compiler improvements that came about over the past year. At least for miniBUDE, the upgrade to GCC 14 yields nice benefits for all of the tested Linux distributions. Keep in mind all of my other Ryzen 9000 series Linux benchmarks to date were on Ubuntu 24.04 LTS with GCC 13, so this is additional performance gains to find compared to the earlier reviews.

NAMD benchmark with settings of Input: ATPase with 327,506 Atoms. Fedora Workstation 40 was the fastest.
NAMD benchmark with settings of Input: STMV with 1,066,628 Atoms. Fedora Workstation 40 was the fastest.

When relying on the NAMD 3.0 binaries across the various Linux distributions, there wasn't much of a measurable performance difference with all of these Linux distributions on recent kernel versions, etc.

SPECFEM3D benchmark with settings of Model: Mount St. Helens. Clear Linux 42170 was the fastest.
SPECFEM3D benchmark with settings of Model: Layered Halfspace. Ubuntu 24.04 LTS was the fastest.
SPECFEM3D benchmark with settings of Model: Tomographic Model. Clear Linux 42170 was the fastest.
SPECFEM3D benchmark with settings of Model: Homogeneous Halfspace. Clear Linux 42170 was the fastest.
SPECFEM3D benchmark with settings of Model: Water-layered Halfspace. Clear Linux 42170 was the fastest.

When firing up SPECFEM3D for this demanding workload, Intel's Clear Linux began to stretch its legs in showing greater performance than the rest thanks to its aggressive compiler defaults and other improvements out-of-the-box.

Related Articles