Every criterion, threshold and gate lives in a versioned rule file. Changing the rules creates a new immutable version so historical ratings stay reproducible.
Start here
Most tier lists tell you the answer. This one shows the working.
Three rules separate a useful rating from a badge: published criteria, evidence that caps ambition, and a full report for every unit.
- Revisions
- 1,079
- Model names
- 898
- Manufacturers
- 46
- Platforms
- 413
The three rules
How this differs from a list that only publishes a letter
Missing laboratory data lowers confidence and caps the attainable tier. The site will not invent a measurement you do not have.
The unit page shows each criterion, its status, the evidence sentence and which gate cleared or blocked the result.
Try it
Why no unit reaches Tier A on specifications alone
Highest attainable
B+
- A+85% and up
- A75% and up
- B+65% and up
- B-50% and up
- C35% and up
- Dbelow 35%
Two opinions, shown side by side
Engine versus published
Click any bar to open that slice of the catalogue.
Now use it
Four ways into the same data
- 1
Browse the catalogue
Filter by tier, efficiency, form factor and more. Every card links to the full report.
Open the catalogue → - 2
Size a build
The calculator turns parts into a capacity and a shortlist with reasons.
Find a unit → - 3
Compare side by side
Two to four units, differences highlighted, best value starred.
Compare → - 4
Read the rules
The criteria and gates on this page are what the engine actually runs.
Stay on this page →
Make a habit of opening the report. The letter is a summary; the criteria are the substance.
Reference
How a rating is produced
The published rules are the engine. Every rating on this site can be opened into the criteria that produced it.
Evidence decides the ceiling
A datasheet alone cannot buy a top tier.
Confidence is computed from specification coverage, attribution, source trust and laboratory evidence. The first matching row below is the highest tier the engine will claim:
- confidence ≥ 85%maximum A+
- confidence ≥ 75%maximum A
- confidence ≥ 65%maximum B+
- confidence ≥ 50%maximum B-
- confidence ≥ 35%maximum C
- confidence ≥ 0%maximum D
Without independent lab data, no unit reaches Tier A here — however strong the architecture looks on paper.
Criteria
22 · v1.0.0
Architecture
- Primary conversion stagevalue_grade · weight 3
- Secondary rectificationvalue_grade · weight 2
- Minor rail regulationvalue_grade · weight 3
- Active power factor correctionflag · weight 2
Standards
- ATX specification generationvalue_grade · weight 2
- AC input rangevalue_grade · weight 2
Efficiency
- Efficiency certificationrank_min · weight 2
Components
- Electrolytic capacitor temperature ratingflag · weight 2
- Fan bearing typevalue_grade · weight 1.5
Protections
- Protection coverageset_coverage · weight 3
- Over-current protection declaredflag · weight 2
- Over-temperature protection declaredflag · weight 1.5
- Protection trip pointsflag · weight 2
Electrical performance
- Measured electrical behaviourflag · weight 3
- +12V ripplenumeric_max · weight 3
- Voltage regulationnumeric_max · weight 2
- +12V transient responsenumeric_max · weight 2
Evidence
- OEM attributionpresence · weight 1.5
- Platform attributionpresence · weight 1.5
- Independent laboratory evidenceflag · weight 2
Warranty and support
- Manufacturer warrantynumeric_min · weight 1.5
Safety
- Safety and integrity recordflag · weight 3
Tier gates
Tier A - reference classspec 12.9
18 requirements
Tier B - strong mainstreamspec 12.10
13 requirements
Tier C - adequatespec 12.11
7 requirements
Tier D - basicspec 12.12
7 requirements
Immediate Tier F
- — No active PFC - the unit cannot satisfy a modern ATX power supply role
- — Confirmed safety hazard, mass failures or falsified certification