This article explains how to tune a server for reproducible benchmarking by reducing run-to-run noise. It demonstrates techniques including pinning to a specific core and locking CPU frequency to achieve deterministic results, reducing coefficient of variation from 2.72% to 1.06% and lower.