Linux 5.0 To Linux 5.9 Kernel Benchmarking Was A Bumpy Ride With New Regressions

Written by Michael Larabel in Software on 9 September 2020 at 12:25 PM EDT. Page 1 of 7. 49 Comments.

Recently carrying out some benchmarks of all major kernel releases from Linux 5.0 through Linux 5.9 ended up yielding some surprising performance changes with the in-development 5.9 kernel. Here's details on this historical look at the kernel performance and what's going on with the Linux 5.9 kernel slowdowns.

This round of Linux 5.0 through Linux 5.9 kernel benchmarking was done on an AMD EPYC 7702 1P server with ASRockRack EPYCD8 motherboard, 128GB of RAM, and 3.8TB Micron 9300 series NVMe SSD. The EPYC 7702 was chosen for its sheer speed especially when it comes to bisecting the kernel with the regressions/changes ultimately uncovered.

The EPYC 7702 is a 64 core / 128 thread part with 2.0GHz base frequency and 3.35GHz boost frequency with 256MB L3 cache. The EPYC 7702 like the rest of the high-end AMD EPYC 7002 processors can build the Linux kernel very quickly.

For the first part of the testing though was using the Ubuntu Mainline Kernel PPA for Linux 5.0 through Linux 5.9-rc4 testing with each of the major kernel releases. The Mainline PPA was used for having the kernel binaries publicly available and the publicly documented kernel configuration files, etc. The bisecting later on was done based off these standard Kconfig files. No hardware changes during testing (obviously) but just swapping out the kernels under test on this Ubuntu 20.04 server.

This comes after some very time intensive tests given all the workloads being evaluated and also exploring the bigger regressions. If you find this testing useful consider showing your support by joining Phoronix Premium to enjoy the site ad-free and multi-page articles such as this on a single page. PayPal tips are also accepted or at the very least to please not use any ad-blocker on this site.

Related Articles