Now, with the CPUFreq solution landing this week in Linux Git, the Linux 5.11 core in its near-final state is looking in great shape for AMD Zen 2/3 hardware on Ryzen laptops and desktops via EPYC servers. The Linux 5.11 development kernel has regressed for most of the last two months, but now that the frequency invariance regression is addressed, not only has the regression disappeared, but it generally has a much better performance compared to with previous versions of the kernel.
Continuing in the December benchmark range, when you first encounter the regression on Zen 2 / Zen 3 when using the default Schedutil governor, here are my latest benchmarks on AMD Linux 5.11 performance now, when the solution was reached by this week’s CPUFreq change and stable Linux 5.11 should be available on Sunday.
As for the desktop, we ran some new benchmarks for the Linux kernel with the Ryzen 9 5900X desktop at stock speeds, the ASUS ROG CROSSHAIR VIII HERO motherboard, 2 x 8 GB DDR4-3600 memory, 1 TB Samsung 980 PRO NVMe SSD and graphics Radeon RX 5600 XT. Since installing Ubuntu 20.10 I tested Linux 5.9.16, Linux 5.10.12 and Linux 5.11 Git as the latest three stable kernel series at that time. Ubuntu Mainline Kernel PPA was used to download these versions of the vanilla kernel in an easily reproducible manner. The default CPUFreq governor on recent AMD hardware kernels is Schedutil.
On the pages that follow prominent individual landmarks, while moving directly to the geometric mean of the 63 landmarks covered in this round.
Linux 5.9 / 5.10 worked roughly the same overall, while with this final actual state of Linux 5.11, performance increased by 5.8% in the range of 63 different test cases. Not bad because performance has been lower in the last two months due to regression, but fortunately now everything is in good shape if the default Schedutil governor is used (as shown in previous tests, the ondemand governor is, also better with Linux 5.11).
The Linux 5.11 kernel came out on top on this Ryen 9 5900X box for 60% of the reference value. In cases where 5.9 / 5.10 won, it was often with thin edges.
Or in last place, the Linux 5.11 Git kernel was in last place for only 5 of the 63 benchmarks. So things are looking pretty good at the moment, Linux 5.11 offering a better performance than previous kernels …
We also ran new benchmarks on the AMD EPYC 7F72 2P server with the Supermicro H11Dsi-NT v2.0 motherboard, 16 x 8 GB DDR4-3200 memory and 1TB Western Digital BLACK SN850 NVMe SSD. Ubuntu 20.10 runs on this AMD EPYC Zen 2 server while the Linux kernels 5.9.16, 5.10.13 and 5.11 Git were used as the latest to initiate these tests.
42 tests were performed on this EPYC server on the latest three kernel series, focusing more on server / workstation loads. With this set of benchmarks, Linux 5.11 based on the geometric mean of all benchmarks came out 6.0% ahead … So quite similar to 5.8% with 5900X, even with the difference in tests. Meanwhile, it was the leader in 76% of the benchmarks compared to 5.10 / 5.9 and was on the last position for only 11% of the tests.
Let’s take a closer look at the performance of Ryzen 9 5900X and EPYC 7F72 2P with Linux 5.11, which should be the version that powers Ubuntu 21.04 this spring.