Run-to-run variance
Also searched as: run to run variance · run-to-run variance · variance · variability
What it means
The normal differences you get when you run the exact same benchmark twice. Nothing's broken — temperatures, background tasks, and luck all shift a little each time.
More detail
It's why a single before-and-after run proves almost nothing; the second run might be higher purely by chance. Serious testing runs a few passes and looks at the spread, not one number. Variance is what sets the noise floor that a real improvement has to beat.
The Compy Blackbook holds 174 graded entries like this one, judged against your hardware — and every change it does make is reversible, with the receipt to prove it.
See Compy →