Rating

The rating shows published results of compatible runs and lets you compare your computer with others. The place in the table depends on the selected profile, engine version, tests and completeness of the record. How the score itself is produced is described in score calculation; the measurement conditions are in the methodology.

Comparison scopes

A position in the rating is not counted only “among everyone”. Available scopes:

Scope	Who your result is compared with
Global	All eligible rating results
Same CPU	Runs with the same CPU
Same GPU	Runs with the same GPU
Same CPU and GPU	Runs where both components match

Comparison in a narrow scope is fairer for evaluating an upgrade of a single component: the global position primarily reflects the class of the whole system. If the CPU or GPU is not recognized in the submitted data, or there are no other eligible results in the scope, comparison in that scope will be unavailable.

Position and percentile

Position is the place of your result among the eligible results of the selected scope: the smaller the number, the higher the result. The percentile shows what share of results in the same scope your result surpasses: 90 means the result is higher than roughly 90% of the compared runs. Both numbers relate only to the selected scope and change together with it. For your own published result, the rating shows the position and percentile separately.

Ranking rule

Only runs of the Candidate profile are ranked — with the canonical durations of all seven tests. Runs of the Quick, Extended and Custom profiles are diagnostic and do not enter the table: their durations and test composition differ, so a direct comparison of scores would be incorrect. Cancelled and incomplete runs are also not ranked.

Composition of a table row

Each rating row contains:

• the position and the name chosen at publication;
• the final PC Score;
• the components Performance, Core Latency Score and Consistency — a breakdown of what exactly is stronger or weaker in this system;
• information about the computer: CPU, GPU and RAM class;
• the classification of run quality;
• the date and percentile.

It is worth looking not only at the final score: a similar PC Score for two systems may add up from a different ratio of speed, latency and consistency. The composition and weights of the components are described in score calculation.

What is sent along with the result

The submitted result contains the category of PC information needed for the scopes and table rows:

• CPU: model, number of physical cores and logical processors;
• GPU: model, amount of video memory and driver version;
• operating system: version and build;
• amount of installed memory.

Submitting a result is voluntary. Check which information about the computer will be published, and do not add personal data to fields available to other users.

Result compatibility

Only results compatible with the current protocol version enter the table. Compatibility is determined by the result file: it records the workload version — a permanent identifier of its composition — and the scoring model, as well as the statistics version and the reference. Runs with an older workload or a different scoring model are incompatible: they are not compared with current results and are not ranked.

How to read the table

1. Choose the scope for your question: the overall class of the system — global, the effect of a single component — a narrow scope.
2. Compare rows with the same methodology: only Candidate runs are ranked, but the run profile is worth checking against the date as well, if the engine has been updated.
3. Look at the components, not only at the final score: they show what the system earned its place through.
4. Remember the limits: a place in the table does not determine FPS in a specific game, does not prove the quality of optimization and does not replace hardware certification.

Before publishing

1. Wait for the test to finish and check the saved result.
2. Make sure the profile and engine version are supported.
3. Check which information about the computer will be sent along with the result.
4. Publish the results of your tests. Do not add personal data to fields available to other users.

Compare in the rating results obtained by the same methodology. How to compare your own runs with each other and separate a change from ordinary variance is described in run comparison.

Last check: 2026-09-20.