Fantastic Bottlenecks and Where to Find Them
summary
Shows concrete fixes and measurable wins: use soft references and default small textures to avoid accidental hard loads, avoid non‑power‑of‑two textures (one extra pixel can blow memory ~10x), and prefer proper texture groups. Occlusion testing of 35,000 queries costs ~13 ms; a movable light with 55,000 unit radius makes its shadow ~20 ms; stationary lighting can be almost twice slower than full dynamic. Use instanced Static Meshes carefully (merging kills culling), simplify collisions to halve frame time, and disable physics on far cars to cut physics time ~5x.
AI-written summary · about
chapters (20)
- 0:00Intro
- 0:22Introduction Ports
- 2:41Common bottlenecks bestiary
- 4:39Assets references
- 6:40Non Power of Two Textures
- 8:11Texture compression types - Choose compression type for the job Default compression
- 9:07Shader Permutations
- 11:09Culling
- 15:11Material Slots
- 16:22Texel Density
- 19:37Draw Calls
- 23:23Simplified Materials
- 24:24Hide-and-Seek
- 27:10Photoreal fans. Lighting
- 29:24Photoreal fans. Materials
- 33:57Photoreal fans. Textures DONTS
- 35:15Sense of Depth
- 36:07Collision complexity 83 ms
- 36:44Physics LODS
- 37:29Overlap Tests
related talks
01h 00mUnreal Engine Game Optimization on a Budget
37mMake your art look good
21mOptimization: a broad intro
28mRuntime optimization patterns - from games to your application
41mBeyond OOP: A Journey From Cache Misses to MASS Hits
47mOptimising for the Web
01h 03mMaximizing Your Game's Performance in Unreal Engine
47mA Thousand Cuts Down to 5