How is the fastest suite decided?

Suites are sorted by mean sample value ascending, so the suite with the lowest average is rank 1; every other suite's delta and factor are calculated relative to that suite's mean.

What does the factor column, like ×1.4, mean?

It is that suite's mean divided by the fastest suite's mean, so ×1.4 means that suite took 1.4 times as long on average as the best-performing suite in the comparison.

Can I export the comparison for a pull request or report?

Yes, use "Copy markdown table" for a ready-to-paste GitHub-flavored table, or "Copy bullet list" for a shorter plain-text summary, both including rank, mean, variance and the delta from the best suite.

How many suites or samples can I compare?

There is no hard limit in the tool itself; add as many suites and samples as you want, though a suite with zero valid numeric samples is dropped from the ranking rather than shown with empty stats.

Benchmark Builder

Compare timing samples across suites by mean and variance, then export the results, locally.

Local, Runs entirely in your browser, your data never leaves your device.

Loading tool…

Rank benchmark suites by mean, variance and speed factor

When you run the same workload under different settings, engines or hardware, the raw numbers alone rarely tell you which one actually won. This tool takes named suites of numeric samples, computes the mean and variance for each, ranks them fastest-first by mean, and shows how far behind the leader every other suite falls, both as an absolute delta and a multiplication factor. All the arithmetic runs in your browser as you type.

How to use the Benchmark Builder

Rename each suite and enter its timing samples, one number per field.
Click "+ Measure" to add another sample to a suite, or the × to remove one.
Click "+ Add suite" to compare more than two sets of results at once.
Set an optional unit label, such as ms, so it appears next to the mean.
Copy the ranked results as a Markdown table or a bullet list once you're happy with the numbers.

Ranking sorts suites by mean ascending, so the fastest suite is rank 1 with a factor of ×1; every other suite shows how many times slower it is and its absolute delta from the best mean, rounded to three decimals.

Variance is the population variance of each suite's own samples, not a comparison against another suite, so a suite with high variance had inconsistent runs even if its mean looks competitive.

Frequently asked questions

How is the fastest suite decided?: Suites are sorted by mean sample value ascending, so the suite with the lowest average is rank 1; every other suite's delta and factor are calculated relative to that suite's mean.
What does the factor column, like ×1.4, mean?: It is that suite's mean divided by the fastest suite's mean, so ×1.4 means that suite took 1.4 times as long on average as the best-performing suite in the comparison.
Can I export the comparison for a pull request or report?: Yes, use "Copy markdown table" for a ready-to-paste GitHub-flavored table, or "Copy bullet list" for a shorter plain-text summary, both including rank, mean, variance and the delta from the best suite.
How many suites or samples can I compare?: There is no hard limit in the tool itself; add as many suites and samples as you want, though a suite with zero valid numeric samples is dropped from the ranking rather than shown with empty stats.