How FrameCheck measures.
FrameCheck turns reviewed, real-world game tests into a clear performance expectation. This page explains what counts, how results are combined, and where uncertainty remains.
Accepted test methods
We prefer a game’s built-in benchmark because it is the easiest route for another tester to reproduce. When no built-in benchmark exists, we accept a repeatable gameplay route lasting at least 10 minutes.
- The CPU and GPU must use stock settings—no overclocking or undervolting.
- V-Sync and frame-rate caps must be disabled.
- The tester reports average rendered FPS and may include 1% low FPS.
- A screenshot, video, result file, or public evidence link must support the run.
What each result records
A result is only matched to the configuration it actually tested: CPU, GPU, memory configuration, game, resolution, quality preset, ray tracing, upscaling mode, and frame generation state.
Frames produced by the game before generated frames are added.
The observed output when frame generation is enabled.
A useful indicator of the slower moments in a run, when supplied.
Review and publication
Community submissions enter a private queue. An administrator checks the evidence, system details, settings, test method, duration, and values before approving or rejecting a result. Uploaded evidence remains private and is only available to administrators.
Approved results are marked as verified and become eligible for estimator matching. A moderation history is retained so a published result can be reopened and removed from estimates if a problem is discovered later.
How an estimate is presented
FrameCheck favors exact verified matches. When several valid samples exist for the same configuration, the estimator uses the median so one unusually high or low run has less influence.
- High confidence: at least 3 samples with a spread of 15% or less.
- Moderate confidence: at least 2 samples with a spread of 25% or less.
- Limited confidence: fewer samples or wider disagreement.
The result card identifies whether a number is measured, sourced, or estimated and shows the available sample count and confidence.
What the number cannot promise
Performance can still change with background software, cooling, power limits, operating-system behavior, game updates, drivers, and differences between scenes. FrameCheck intentionally avoids asking users to choose every driver or game version so the estimator stays approachable.
Treat every FPS figure as a practical expectation, not a guarantee. Where no matching verified result exists, FrameCheck says so instead of presenting placeholder data as fact.