Methodology
How Specora actually decides what's limiting your performance, and where it draws the line at “we don't know.”
Why not a single bottleneck percentage?
A “23% bottleneck” implies one universal number describes your whole PC, in every game, at every setting. That isn't how bottlenecks work. The same CPU and GPU pairing can be GPU-limited in one game at 4K Ultra and CPU-limited in a different, simulation-heavy game at 1080p. Specora always evaluates a specific combination: your CPU, your GPU, your RAM, the resolution, the graphics preset, your target framerate, and the game's own workload profile.
How the bottleneck call is made
Each CPU and GPU has a gaming performance index derived from its specification tier. The engine scales each component's effective score against how demanding the chosen resolution, preset and specific game are for that component type. Whichever component has less headroom relative to what the scenario demands is reported as the likely limiter. This comparison happens entirely in a deterministic rules engine — no AI model is asked to guess or invent this figure.
The confidence system
Every performance estimate carries a confidence tier — High, Medium, Low, or Insufficient — with a stated reason. High confidence means a real logged benchmark matched exactly. Medium or Low means the estimate is derived from relative hardware indices rather than a direct measurement. Insufficient means the combination falls outside what Specora can reliably approximate, in which case it says so rather than guessing.
Game catalog vs. benchmark coverage
A game being searchable in Specora means the catalog knows it exists — nothing more. Whether a precise FPS estimate exists for your exact hardware depends entirely on separate benchmark data. Specora will never treat catalog size as a substitute for measured performance data.
Reporting inaccurate results
Every result has a “Was this analysis accurate?” control and a way to report a specific issue. Reports and feedback are stored separately from benchmark data and reviewed manually — user feedback never automatically becomes “verified” performance data.