A practical Game Thread optimization checklist for UE5

When a UE5 title targets everything from PS5 down to Switch-class hardware, the Game Thread is usually the first budget you blow. This is the checklist I work through before touching any code.

1. Measure before guessing

• stat unit — confirm the Game Thread (not Draw or GPU) is actually the bottleneck.
• stat game — break down tick, animation, and physics cost.
• Unreal Insights with -trace=default,task — find the exact frames that spike and what ran in them.

If the spike only happens on device, profile on device. Editor numbers lie.

2. The usual suspects

1. Too many ticking actors. Search for components that tick every frame but only need event-driven updates. PrimaryActorTick.bCanEverTick = false is the cheapest optimization in the engine.
2. Tick frequency. For things that must tick, SetTickInterval at 10–15 Hz is often indistinguishable from 60 Hz for non-player-facing logic.
3. Blueprint hot paths. A Blueprint tick doing per-frame math is an easy nativization candidate — move the inner loop to C++ and keep the Blueprint as configuration.
4. Spawning during gameplay. Pool projectiles, VFX actors, and UI widgets. SpawnActor mid-combat is a frame-time grenade.
5. Overlap and trace volume. Collapse per-frame traces into batched or asynchronous queries (AsyncLineTraceByChannel) where the result can arrive a frame late.

3. Verify on the weakest target

A change that wins 0.3 ms on PC can be noise on Switch, and vice versa — CPU cache behavior and core counts differ completely. Re-run the same Insights capture on the weakest platform after every batch of changes, and keep a before/after table in the PR description so the win is verifiable.

The pattern behind all of this: make the frame do less work, make remaining work event-driven, and never trust a number you didn't measure on target hardware.