Comparisons

Use these pages to see how Onity’s pillars compare with the libraries they replace, feature by feature and with measured results. Each page keeps every number next to its conditions (date, Unity version, scripting backend, process count, settings) and states what the measurement does not cover. This hub has no numbers; the pages do.

What is compared

Page Onity pillar Compared with What it measures
DI vs VContainer and Zenject OnityContainer, OnityObjectPool<T>, PooledFactory VContainer, Zenject Feature breadth, resolve and build timing, allocation per operation, pooled factory timing
Reactive vs R3 and UniRx Onity.Reactive R3, UniRx Feature breadth and the measured operator workloads
OnityTask vs UniTask OnityTask UniTask The Release Player gate, API coverage and the remaining gaps
Messaging vs MessagePipe Onity.Messaging MessagePipe Feature breadth and the measured publish, subscribe and dispose, keyed and async workloads

How to read a ratio

  • A ratio is Onity’s time divided by the other library’s time for the same workload in the same process. Below 1 means Onity took less time; above 1 means it took more.
  • A page reports the median across the measured processes and names the worst process, so a single fast or slow run cannot carry a claim on its own.
  • Timing results are timing results. Allocation, other platforms and other workloads are separate measurements, and a page says when it has none.
  • The conditions travel with the number: the Unity version, the backend (Mono or IL2CPP), the number of processes, the date and the settings that differ from the defaults. The known slower cases are stated on the same page, not elsewhere.
  • Every result was produced by the runners in this repository on one Windows machine. They are reproducible with those runners; they are not a guarantee for other hardware, Unity versions or object graphs.

Evidence pages

The comparison pages cite these records, which are kept unchanged as measured:

Older measurements that the current pages no longer carry are in Measurement history, unchanged and dated.


Table of contents