AMD Ryzen "Renoir" CPU Frequency Scaling Governor Performance

Written by Michael Larabel in Processors on 22 July 2020. Page 1 of 7. 12 Comments

With 129 tests carried out while also looking at the CPU power consumption and temperatures during benchmarking, here is a look at how the CPU frequency scaling governor plays a role in the performance of the latest-generation AMD Ryzen 4000 "Renoir" laptops for Linux.

Running most Linux distributions on an AMD Ryzen 4000 series laptop will result in the CPUFreq "ondemand" governor used by default as a sane default for controlling the CPU frequency scaling behavior. But there is also the performance governor as the most performance-minded and quickest to ramp up CPU frequencies when needed, conservative for more conservative frequency scaling decisions, schedutil as the possible future default that leverages the Linux kernel's scheduler utilization data for making more accurate power management decisions, and powersave as an attempt to conserve power use over performance. Those different CPUfreq governors are being benchmarked in this article with the Lenovo Flex 5 with Ryzen 5 4500U processor.

There is also the AMD CPPC CPUFreq driver that was initially published last year for making use of Collaborative Processor Performance Controls with Zen 2. But that driver has yet to be mainline and in fact haven't seen any new revisions to it in recent months. Hopefully AMD engineers will get back to working on that CPPC CPUFreq driver soon for potentially better power/performance handling on Linux systems.

The LENOVO LNVNB161216 with Ryzen 5 4500U was running an Ubuntu 20.10 development build for the bleeding edge experience. Additionally, the system was upgraded to a Linux 5.8 Git snapshot for AMD Renoir CPU temperature monitoring as well as the new AMD Energy driver for being able to monitor the real-time power consumption of the CPU package under Linux.

AMD Ryzen 5 4500U Governors

Via the Phoronix Test Suite 129 different benchmarks were carried out while also monitoring the CPU package consumption and generating the performance-per-Watt on a per-test basis and the per-test CPU temperature as well as the Vega onboard graphics core temperature too. The only change between testing was changing out the CPUFreq governor being utilized.

Related Articles