Benchmarking The Performance Impact To AMD Inception Mitigations

Written by Michael Larabel in Software on 15 August 2023 at 10:30 AM EDT. Page 4 of 8. 35 Comments.
Numpy Benchmark benchmark with settings of . off was the fastest.

Numpy also recorded a hit from the new mitigations.

ClickHouse benchmark with settings of 100M Rows Hits Dataset, First Run / Cold Cache. off was the fastest.
ClickHouse benchmark with settings of 100M Rows Hits Dataset, Second Run. off was the fastest.
Apache Spark benchmark with settings of Row Count: 1000000, Partitions: 100, Group By Test Time. off was the fastest.
Apache Spark benchmark with settings of Row Count: 1000000, Partitions: 100, Inner Join Test Time. off was the fastest.
CockroachDB benchmark with settings of Workload: KV, 50% Reads, Concurrency: 128. off was the fastest.
CockroachDB benchmark with settings of Workload: KV, 95% Reads, Concurrency: 128. off was the fastest.
Apache IoTDB benchmark with settings of Device Count: 200, Batch Size Per Write: 1, Sensor Count: 200. off was the fastest.
Apache IoTDB benchmark with settings of Device Count: 500, Batch Size Per Write: 1, Sensor Count: 200. IBPB was the fastest.
Apache IoTDB benchmark with settings of Device Count: 200, Batch Size Per Write: 100, Sensor Count: 500. off was the fastest.

Various database solutions were showing varying degrees of performance penalties from using the AMD Inception mitigation.

MariaDB benchmark with settings of Clients: 4096. off was the fastest.
MariaDB benchmark with settings of Clients: 8192. off was the fastest.

MariaDB saw some of the most significant performance penalties from the new kernel changes with safe RET.

PostgreSQL benchmark with settings of Scaling Factor: 100, Clients: 800, Mode: Read Only. off was the fastest.
PostgreSQL benchmark with settings of Scaling Factor: 100, Clients: 800, Mode: Read Write. off was the fastest.

The PostgreSQL database server also recorded some slowdowns as a result of the mitigated kernel.

Related Articles
Linux 6.13 Features: AutoFDO+Propeller Optimizations, Many AMD Additions & SDUC + NVMe 2.1 Support
AMD BIOS Tuning Guide Impact For Boosting AI/ML Performance On EPYC 9005 Series
New AMD ERAPS Feature Yields Additional Performance Gains On Zen 5
Initial Benchmarks Of The AMD AOCC 5.0 Compiler On 5th Gen EPYC
Linux 6.12 Features Are Super Exciting With Real-Time, Sched_ext, Intel Xe2 & Raspberry Pi 5
AmpereOne Performance With GCC vs. LLVM Clang Compilers