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

Written by Michael Larabel in Software on 6 April 2023 at 04:00 PM EDT. Page 5 of 5. 13 Comments.
Stress-NG benchmark with settings of Test: Poll. Clear Linux: 384 Threads was the fastest.

Not only was Intel's Linux stack delivering much greater performance at 192 cores for the PostgreSQL database server, it also translated into rather lofty advantages for testing various kernel primitive operations with Stress-NG.

Stress-NG benchmark with settings of Test: Semaphores. Clear Linux: 384 Threads was the fastest.
Stress-NG benchmark with settings of Test: Context Switching. Clear Linux: 384 Threads was the fastest.

Across the various workloads tested -- which was only limited by my time and having just one AMD Genoa server currently -- for these very time intensive tests, Intel's Linux optimization work continues showing huge advantages at higher core counts... Even for AMD EPYC servers.

RocksDB benchmark with settings of Test: Random Fill. Clear Linux: 96 Threads was the fastest.
RocksDB benchmark with settings of Test: Random Read. Clear Linux: 192 Threads was the fastest.
RocksDB benchmark with settings of Test: Read Random Write Random. Clear Linux: 96 Threads was the fastest.

With Meta's RocksDB database software, more Intel's optimized Linux kernel continued showing advantages compared to CentOS Stream and the current state of Ubuntu Linux.

Geometric Mean Of All Test Results benchmark with settings of Result Composite, Intel Kernel Scaling Optimizations For AMD Genoa. Clear Linux was the fastest.

Similar to the recent Intel Xeon Scalable Sapphire Rapids scaling benchmarks, AMD 4th Gen EPYC "Genoa" showed much benefit from Intel's open-source optimizations too up to the 384 thread configuration of two flagship EPYC 9654 processors. Intel's Linux optimizations over the year have proven extremely beneficial to the entire ecosystem and in many instances do end up benefiting AMD x86_64 hardware too. As next year's Sierra Forest launch approaches and higher core counts with next-generation Xeon Scalable processors, it's likely -- and hoped -- we'll see much more out of Intel's excellent software engineering team for uncovering and addressing more low-level kernel contention issues and other scalability challenges. With AMD's growing success in the data center, hopefully their Linux software team continues to grow as well and we see more generic kernel optimizations/improvements from them too.

As these latest Intel patches continue undergoing review for upstreaming and any other interesting performance work published, stay tuned to Phoronix to learn about these interesting happenings.

If you enjoyed this article consider joining Phoronix Premium to view this site ad-free, multi-page articles on a single page, and other benefits. PayPal or Stripe tips are also graciously accepted. Thanks for your support.


Related Articles
About The Author
Michael Larabel

Michael Larabel is the principal author of Phoronix.com and founded the site in 2004 with a focus on enriching the Linux hardware experience. Michael has written more than 20,000 articles covering the state of Linux hardware support, Linux performance, graphics drivers, and other topics. Michael is also the lead developer of the Phoronix Test Suite, Phoromatic, and OpenBenchmarking.org automated benchmarking software. He can be followed via Twitter, LinkedIn, or contacted via MichaelLarabel.com.