Make it fast
Performance
How geometry, textures, draw calls, and delivery choices affect real-world speed.
10 articles
Why Large GLB Files Slow Websites
Trace a GLB through download, decode, parsing, image expansion, GPU upload, compilation, and rendering to fix the stage users actually feel.
How to Diagnose a Slow 3D Model
Use a repeatable decision tree to isolate network, decode, memory, CPU submission, vertex, fragment, shader, and lifecycle bottlenecks.
How to Build Mobile 3D Performance Budgets
Turn device tiers, frame deadlines, memory limits, network conditions, thermal behavior, and scene measurements into enforceable mobile 3D budgets.
3D Load Time vs. Runtime Performance
Separate startup acquisition and preparation from recurring CPU and GPU frame work, then verify that each optimization improves the intended experience.
Meshopt vs. Draco vs. Quantization for glTF
Compare glTF quantization, Draco, and meshopt by precision, transfer bytes, decode behavior, runtime layout, compatibility, and deployment cost.
Why Draco Compression Is Not Enough for a Fast GLB
Learn exactly what Draco changes, what remains expensive after decode, and how to pair geometry compression with texture, scene, and runtime optimization.
GPU Memory for 3D Assets: A Practical Budgeting Guide
Estimate textures, geometry, render targets, shadows, transient allocations, and duplicated resources before a real-time 3D scene exhausts device memory.
Draw Calls: The Hidden Cost Inside a 3D Model
Understand how primitives, materials, passes, instancing, transparency, batching, and culling turn one asset into many rendering submissions.
Texture Size vs. Mesh Size: Finding the Real 3D Bottleneck
Compare textures and mesh data across transfer, decode, GPU memory, bandwidth, and frame time to choose the right optimization first.
Triangle Count vs. File Size: Why the Numbers Do Not Move Together
Calculate how indices, rendered vertices, attributes, textures, animation, and compression contribute to GLB size before choosing decimation.