Intel Linux Optimizations Help AMD EPYC "Genoa" Improve Scaling To 384 Threads

Last month I wrote about Intel's Linux kernel engineering improvements to help enhance CPU scaling across various workloads by addressing low-level bottlenecks within the kernel. It's an area we'll likely see Intel continue to invest in as Sierra Forest comes next year with 144 E cores per socket. Already with the Linux kernel patches Intel is carrying at the moment via their in-house distribution, there are some significant benefits for Xeon Scalable Sapphire Rapids. I was curious to see how this Intel-focused work impacted AMD EPYC servers and thus in today's article is a similar analysis using two AMD EPYC 9654 "Genoa" flagship processors while evaluating Intel's Linux kernel optimizations.

Today's testing is focused on how well AMD EPYC Genoa is scaling under Linux with various workloads and the impact of Intel's ongoing Linux kernel low-level improvements for addressing various bottlenecks, contention issues, and other inefficiencies within core parts of the kernel. As Intel's kernel optimization work is ongoing, for easily assessing the current state of affairs their Clear Linux distribution was used with their latest kernel where they continue to carry their various out-of-tree patches that have yet to be reviewed and upstreamed into the mainline Linux kernel. Intel engineers continue working on upstreaming their relevant kernel changes -- and as a company continue to be one of the largest contributors to the upstream kernel -- while for testing the immediate impact of this recent work it's easiest to do so on their Clear Linux platform.

The latest Clear Linux performance was then compared against Ubuntu 23.04 in its latest development state (for having Linux 6.2 and the very up-to-date packages there rather than the year-old Ubuntu 22.04 LTS state) as well as also testing CentOS Stream 9 for that latest RHEL9 state.

All of these Linux distributions were tested on the same AMD Titanite dual-socket server with EPYC 9654 processors providing a combined 192 cores / 384 threads. With the same hardware and clean installs of each Linux OS, the following core/thread configurations were tested for this AMD 4th Gen EPYC server:

- 3 cores
- 6 cores
- 12 cores
- 24 cores
- 48 cores
- 96 cores
- 192 cores
- 384 threads (192 cores + SMT; fully-enabled EPYC 9654 2P configuration)

Like with the prior round of Sapphire Rapids testing, this allows seeing in turn how Intel's optimized kernel is performing relative to the Linux 6.2 kernel being shipped by Canonical with the upcoming Ubuntu Server 23.04 as well as the current kernel state with RHEL9 / CentOS Stream 9.