WebGPU: Next-generation 3D Graphics on the Web
summary
Claims concrete speedups: precomputing precompiled render bundles cuts CPU draw calls by an order of magnitude in complex scenes, and GPU compute shaders for matrix multiply runs 512×512 in milliseconds versus seconds on CPU. Implementation uses native libraries (Dawn and WGPU) and a WebAssembly toolchain to port a native aquarium benchmark to the web. Demos include a Babylon.js forest showing steady FPS on WebGPU and a working matrix demo; some prototypes are still fragile.
AI-written summary · about
chapters (15)
- 0:00Intro
- 2:19GPU architecture
- 9:48Scene Graph in WebGL
- 15:15Demo of command encoders.
- 17:42GPU are very good at parallel processing
- 20:11GPU compute 101: matrix multiplication
- 21:21Initialize WebGPU
- 22:322. Allocate memory for the matrix data.
- 23:11Create the data "group"
- 23:35a GPU program source.
- 25:13Encode the compute commands.
- 25:51Encode the readback commands.
- 26:25Submit work to the GPU.
- 28:39GPU compute 101 demo!
- 29:41Implementation: Chromium
related talks
34mNext-Generation 3D Graphics on the Web
34mWhat WebGPU means for Web platform?
40mThe WebGPU Revolution: Full GPU Power on the Web and Beyond
40mWebGPU and Babylon.js: a Fruitful Collaboration
22mWebGL For More Than Graphics: Is It Possible?
28mTurbo charging 2D on the web
30mWasi-Webgpu: Build Beautiful Graphics and Safely Run AI with Any GPU
32mGPU.JS - GPU Accelerated JavaScript