Making it perform

Triangle census, collider choice, what actually costs you.

guides

Decals: projection, determinism, and performance in s&box

The practical patterns for Sandbox.Decal: orienting the projection box to a hit normal, keeping a decal byte-identical across multiplayer peers, making paint follow an animated character, and scaling to thousands of decals without a performance cliff, plus the pitfalls that catch each one.

Performance investigation: the measure-first harness

A disciplined method for answering "the game feels slow" in an s&box project without shipping speculative optimizations: environmental pre-checks, measurement tooling, an isolation matrix, and a verdict discipline that prevents wasted work.

Runtime terrain meshing: chunked greedy voxel/heightfield terrain

The method for large runtime-generated terrain in s&box: a persistent cell grid as the single source of truth, chunked greedy meshing (tops + skirts), collision decoupled from render grain, palette-atlas UVs, and dirty-chunk remesh for a live terrain brush.

World scale and coordinate limits: units, the 20k folklore, and float precision from origin

What "a 20k world" actually means (units vs meters vs Source-1 map folklore, and they are ~1550x apart), why s&box has no VBSP-style coordinate wall, and the real modern ceiling: fp32 position precision that coarsens with distance from origin. Measured numbers, the lore separated from the measurements, and the reproduction method.

✓ 26.07.08e

"A frame-budget method for stylized s&box scenes"

Census tris × instances before decimating; prefer BoxCollider over ModelCollider for decorative props.

performanceart-pipeline

✓ 26.07.08e

"Chunked mesh generation is math-bound, not draw-call-bound"

For a chunked runtime-mesh generator, REGEN (single-threaded generation CPU) is the binding resource, not fps or draw calls. Quartering the chunk count produces zero regen change; the lever is threading the passes.

performanceprocgen

✓ 26.07.08e

"Collider choice that actually costs you"

BoxCollider is cheap and follows WorldScale; Capsule/ModelCollider don't scale with WorldScale and ModelCollider is costly on clutter.

performancephysics

✓ 26.07.22

"Editor play mode is hard-capped at 60 fps: cvars won't lift it"

Editor-embedded play mode pins at exactly 60 fps regardless of cvars. The cap is the compositor's vsync on the editor window, not the engine frame sync. Under the cap even the p50 pins to 16.67 ms: read the tail (p99, worst frame, over-budget count), not the median.

physicsperformance

✓ 26.07.22

"Editor vsync hides physics cost: time the step directly"

Frame fps pins at 60 in the editor regardless of physics load: implement IScenePhysicsEvents.PrePhysicsStep/PostPhysicsStep with a Stopwatch to get the true, vsync-free CPU cost of the solver step.

performancephysics

✓ 26.07.22

"FrameStats and PerformanceStats are accessible from game code: no engine benchmark system needed"

Sandbox.Diagnostics.FrameStats and PerformanceStats are whitelist-clean and reachable from game-assembly components: build a relative perf probe without any engine benchmark API. Raw GC.GetAllocatedBytesForCurrentThread/GC.CollectionCount are SB1000-blocked in-editor; use PerformanceStats instead.

whitelistperformance