The engine, opened up
A deep dive into the NBenchmark measurement engine, from the worker protocol and clock probing to optional-stopping correction.
Read the postWriting
Measuring .NET honestly
Deep dives into NBenchmark - adaptive measurement, honest statistics, outlier handling, and the traps that make benchmarking .NET harder than it looks.
When the built-ins are not enough
Extending NBenchmark with custom reporters, outlier detectors, and OpenTelemetry integration for advanced observability.
One codebase, three runtimes
Comparing performance across .NET versions to make data-driven upgrade decisions.
State that survives between benchmarks
Managing instance lifetimes and shared state to prevent contamination and ensure statistically valid results.
Why your numbers come from another process
An exploration of process isolation, runtime profiles, and why measuring in-process is a recipe for misleading results.
When the numbers look wrong
A field guide to diagnosing impossible benchmark results, from clock resolution issues to state contamination.
Fail the build when it gets slower
Turning performance measurements into enforceable gates using CLI regression limits and test-framework integrations.
Cutting the noise at its source
Reducing environmental interference using CPU affinity, process priority, and thread control for maximum measurement precision.
One run cannot tell you how much it moves
Using multiple launches and paired ratios to distinguish between a lucky run and a real performance regression.
Benchmarks that take arguments
Measuring performance as a function of input using parameterized benchmarks and categories to organize scaling sweeps.
Output that travels
Moving beyond the terminal with JSON, Markdown, and CSV reporters to make performance data shareable and archivable.
Counting what the GC sees
How NBenchmark measures allocations and manages GC behavior to balance real-world noise with measurement precision.
When your code outruns the clock
Learn how NBenchmark uses ops-per-sample calibration to accurately measure sub-microsecond operations that finish faster than the system clock can notice.
Page 1 of 2