Web AI leaps forward on Intel AI PCs
summary
WebNN on GPU doubles real-world throughput while NPU coverage expands to run 20 top Whisper/TransformerJS models; Whisper goes from ~43 to ~100 tokens/sec on GPU and ~98 on NPU, and Stable Diffusion drops from ~900ms to ~400ms on GPU (~600ms on NPU). Cross-device buffer reuse yields ~18% CPU‑GPU and ~50% KV‑cache gains; WebGPU matrix multiply and SIMD tweaks give ~2x speedups and tokens/sec from ~15 to ~20. Panther Lake (early 2026) targets 180 TOPS vs Lunar’s 120, with up to 12 XC cores, each core with eight XMX engines, WebNN doubles GPU performance.
AI-written summary · about
chapters (5)
- 0:00Introduction and agenda
- 1:51WebNN progress and demos
- 7:56WebGPU optimizations
- 10:01Intel Panther Lake preview
- 12:29W3C collaboration and conclusion
related talks
30mBeyond the Web of Today
16mThe Web is AI Ready—maximize your AI web development with WebNN
39mThe AI PC and the Era of the Modern Web
41mEscaping the Cloud: High-Performance AI in your Browser
22mBoring but Effective: WebAI That Ships
17mRunning Intelligence at the Edge: Web AI in Action
45mPrivacy-First In-Browser Generative AI Web Apps
24mWeb Apps of the Future With Web AI