01At a glance
- Rig
- Reference hardware per /hardware
- Warm-up
- Same route twice; second is measured
- Evidence
- Build + firmware + driver recorded
- Promotion
- Repeated clean runs, never one
| From | To | Requires |
|---|---|---|
| Test | Beta | Repeated end-to-end runs with documented workarounds |
| Beta | Stable | Repeated full sessions at defaults, no blockers |
| Any | Drop | One confirmed regression on re-verification |
02Normative procedure
- Fix the rig: reference hardware per Minimum and recommended, current drivers, passing Vulkan check and AVX2 probe.
- Fix the software: build from Download Meridian, official PS5UPDATE.PUP via File then Install Firmware, owned dump indexed per Index your library.
- Warm the cache: run the same route, lap, or sortie twice. The second pass is the measurement — see Shader cache.
- Run the session: full sessions for Stable candidacy, end-to-end with notes for Beta, menus plus early gameplay for Test — per Status definitions.
- Record evidence: build number, firmware version, GPU driver, settings, and observed defects. Without evidence the run did not happen.
03Promotion and regression rules
- Promotion requires repeated clean runs across sessions, not one good evening. Single runs promote nothing.
- Regression on re-verification drops the rating immediately and is recorded with the causing build.
- Ratings never transfer across builds silently. Each build re-verifies the Stable titles leaders first.
- Own your library: method runs use owned dumps only. 400+ titles indexed; each earns its rating separately.
04Questions
Q.01How many runs promote a title?
Multiple clean sessions under identical conditions. One clean run is a candidate, never a promotion.
Q.02What evidence must a report include?
Build number, firmware version, GPU driver, settings, and warmed-cache runs. Missing any, unactionable.
Q.03Do faster rigs earn faster promotions?
No. Ratings follow the reference rig so results compare across users and builds.