AMD Ryzen, EPYC 5~6% Faster Out-Of-The-Box With Linux 5.11

Now with the CPUFreq fix landing this week in Linux Git, the mainline Linux 5.11 kernel in its near final state is looking in very good shape for AMD Zen 2/3 hardware from Ryzen laptops and desktops through EPYC servers. The Linux 5.11 development kernel was regressed for the better part of the past two months but now that the frequency invariance regression is addressed, not only is the regression gone but generally is performing much better compared to prior kernel versions.

Continuing in the array of benchmarks since December when first encountering the regression on Zen 2 / Zen 3 when using the default Schedutil governor, here are my final benchmarks looking at the AMD Linux 5.11 performance now that the fix by way of the mentioned CPUFreq change landed this week and Linux 5.11 stable should be out on Sunday.

On the desktop side I ran some fresh Linux kernel benchmarks with the Ryzen 9 5900X desktop at stock speeds, ASUS ROG CROSSHAIR VIII HERO motherboard, 2 x 8GB DDR4-3600 memory, 1TB Samsung 980 PRO NVMe SSD, and Radeon RX 5600 XT graphics. From the Ubuntu 20.10 installation I tested Linux 5.9.16, Linux 5.10.12, and Linux 5.11 Git as the three latest stable kernel series at the time. The Ubuntu Mainline Kernel PPA was used for fetching these vanilla kernel builds in an easily reproducible manner. The default CPUFreq governor on the recent kernels for AMD hardware is Schedutil.

On the pages that follow the prominent individual benchmarks while let's just jump straight to looking at the geometric mean of the 63 benchmarks ran this round.

Linux 5.9/5.10 performed about the same overall while with this effectively final state of Linux 5.11, the performance is up by 5.8% across the span of 63 different test cases. Not bad for the performance being lower much of the past two months due to the regression but thankfully now all is in good shape if using the default Schedutil governor (as shown in prior tests, the ondemand governor is also better off with Linux 5.11 too).

The Linux 5.11 kernel came out in first place on this Ryen 9 5900X box for 60% of the benchmark. In the cases where 5.9/5.10 won, it was often by slim margins.

Or in last place finishes, the Linux 5.11 Git kernel was in last place for just 5 of the 63 benchmarks. So things are looking fairly good at this point with Linux 5.11 delivering a more performant experience than prior kernels...

I also ran fresh benchmarks on the AMD EPYC 7F72 2P server with Supermicro H11Dsi-NT v2.0 motherboard, 16 x 8GB DDR4-3200 memory, and 1TB Western Digital BLACK SN850 NVMe SSD. Ubuntu 20.10 was running on this AMD EPYC Zen 2 server while the Linux 5.9.16, 5.10.13, and 5.11 Git kernels were used as the latest when initiating those tests.

42 tests were carried out on this EPYC server across the three latest kernel series in focusing more on server/workstation loads. With this set of benchmarks, Linux 5.11 based on the geometric mean of all the benchmarks came out 6.0% ahead... So quite similar to the 5.8% with the 5900X even with the difference in tests. It meanwhile was the leader in 76% of the benchmarks compared to 5.10/5.9 and was in last position for just 11% of the tests.

Let's look in more detail at the Ryzen 9 5900X and EPYC 7F72 2P performance with Linux 5.11, which should be the version powering Ubuntu 21.04 this spring.