AMD P-State EPP Performance With EPYC On Linux 6.3

Written by Michael Larabel in Software on 3 April 2023 at 11:30 AM EDT. Page 3 of 6. 9 Comments.
uvg266 benchmark with settings of Video Input: Bosphorus 4K, Video Preset: Medium. acpi_cpufreq ondemand was the fastest.
uvg266 benchmark with settings of Video Input: Bosphorus 4K, Video Preset: Medium. acpi_cpufreq ondemand was the fastest.
uvg266 benchmark with settings of Video Input: Bosphorus 4K, Video Preset: Medium. acpi_cpufreq ondemand was the fastest.

With the uvgg266 open-source H.266 encoder there was a small improvement in delivering the best power efficiency on this AMD Milan-X server when using the powersave governor with balanced performance ePP preference.

OpenVKL benchmark with settings of Benchmark: vklBenchmark ISPC. amd_pstate performance was the fastest.
OpenVKL benchmark with settings of Benchmark: vklBenchmark ISPC. amd_pstate performance was the fastest.
OpenVKL benchmark with settings of Benchmark: vklBenchmark ISPC. amd_pstate performance was the fastest.

For many of the benchmarks conducted there wasn't a whole lot of difference aside from the amd_pstate powersave configuration being much slower and at times amd_pstate schedutil also being slower to improve the power efficiency based on the scheduler utilization data.

7-Zip Compression benchmark with settings of Test: Compression Rating. amd_pstate performance was the fastest.
7-Zip Compression benchmark with settings of Test: Decompression Rating. acpi_cpufreq schedutil was the fastest.
7-Zip Compression benchmark with settings of Test: Decompression Rating. acpi_cpufreq schedutil was the fastest.
7-Zip Compression benchmark with settings of Test: Decompression Rating. acpi_cpufreq schedutil was the fastest.

At least on the Milan-X server side, there wasn't any "game changers" with P-State EPP.


Related Articles