Read diagnostic guide

Open method · Browser-only · No account

Frame-Time Evidence Lab

Test whether a repeat pass changes frame delivery at the same scene. Paste raw frame duration in milliseconds from two comparable captures. The calculations are visible and the report is reproducible.

1. Paste paired capture series

Accepts one number per line or a one-column CSV. Export only the frame-time column from your profiler. Both captures must use the same game build, resolution, settings, route, hardware and driver. Do not clear caches between passes. No data is uploaded.

3. Method and limits

This is an original GAMIC News utility, not a measured hardware benchmark. The loaded example is fictional illustration only. No pass result proves a driver, shader or pipeline cause on its own: asset streaming, CPU scheduling, background jobs, thermal limitations and game-specific behavior may produce similar patterns.

  1. Median / P95 / P99 frame time: nearest-rank percentiles on the sorted positive measurements (lower is smoother for equivalent scenes). P99 uses the observation at ceil(0.99 × n).
  2. Slowest 1% mean frame time: arithmetic mean of the largest max(1,ceil(n/100)) observations; its reciprocal ×1000 estimates a clearly defined slowest-1%-mean FPS proxy, not an industry-standard or benchmark-vendor-defined "1% low".
  3. Spike count: frame duration greater than max(2 × pass median, 1000 ÷ target Hz). This threshold is an explicitly chosen diagnostic heuristic, not a scientific causal classification.
  4. Compare like with like: use captures with similar sample counts and the same route. A shorter scene, pause menu or different gameplay invalidates a simple comparison.

Methods and technical references: Microsoft Direct3D 12 — Managing graphics pipeline state; Epic Games — PSO precaching; GAMIC News — controlled comparison guide.

Data policy: Processing takes place entirely in your browser. No login, cookies, external libraries, storage, analytics, or network requests are used by this tool. The exported JSON contains only the measurements, calculations and declared method. Publication date: 25 October 2026. This is independently authored by GAMIC News; no external lab reviewed or certified these results.