Regression
A regression is a measurable drop in performance versus a previous baseline on the same hardware, caused by a software, driver, firmware, or hardware change.
Also known as performance regression
Measured by SystemCheck
Part of the four-stage run: a 60-second sustained CPU load reporting throughput, decay and consistency; a memory stage reporting bandwidth and latency; a GPU raster load at a fixed resolution reporting frame rate, pacing and decay; and a path tracer reporting samples per second. The paragraphs below say which metric carries this term.
Regressions arrive quietly. A driver update, a Windows feature update, a BIOS revision, a new background security agent, or an accumulating dust layer each cost a few percent, and the machine gets slower over months without any single moment where it obviously broke.
Catching them requires a stored baseline and a repeatable test. The size of the drop matters too. A change smaller than your machine's run-to-run variance is not a regression, it is noise, and chasing it wastes time. Establish variance first, then treat drops beyond it as real.
SystemCheck's saved reports make this practical. Compare CPU throughput, decay, and consistency against a stored baseline, and check the backend and measurement revision fields to be sure you are comparing like with like. A drop concentrated in decay rather than peak throughput points at cooling or power; a drop across the board points at software or driver.
When you find one, narrow it down by undoing changes one at a time in reverse order, re-testing after each. The change whose reversal restores the result is the cause. If nothing you changed restores it, the cause may be physical, such as dust or ageing thermal paste, and a cleaning followed by a fresh test is the next step. Keep every report along the way; the sequence is the evidence.
Related terms