Qualcomm Technologies and the power of Windows AI Q&A
summary
Qualcomm demonstrates running Stable Diffusion entirely on-device on Windows on Snapdragon, showing a live demo of a ~1 billion parameters model and describing steps to reach ~10B on-edge via aggressive quantization, memory partitioning and token batching. Developers get a path from proprietary Neural Processing SDK to ONNX Runtime and direct hardware access through AI Engine Direct with Windows-native tooling, sample notebooks for quantization, and monthly stack updates plus guidance for hybrid cloud/edge deployment.
AI-written summary · about
chapters (23)
- 0:00Qualcomm AI on Windows
- 1:12AI stack and ecosystem
- 2:39ONNX runtime for AI inference
- 4:02AI Engine Direct and acceleration
- 5:29Generative AI at the edge challenges
- 6:37Stable Diffusion demo on Snapdragon
- 8:24Qualcomm’s AI investment vision
- 10:52Edge AI models future limits
- 12:18Developer call to action
- 15:39Windows Studio effects with NPU
- 16:41Q&A
- 17:25ONNX runtime and SDK
- 19:10Snappy SDK improvements
- 20:34Cross-hardware Qualcomm AI
- 21:57Offloading AI to NPU benefits
- 23:23Native ONNX runtime deployment
- 25:10Quantization for NPU hardware
- 26:57Why edge AI models
- 28:23Legacy silicon support
- 30:52Scaling to 10B parameters
- 32:45AI accelerators ubiquity
- 35:33Qualcomm booth AI apps
- 37:24Final demo and closing
related talks
16mRealize your vision with a guide to building Windows on Snapdragon
15mBuilding Gen-AI experiences using LLM and LVM
05mDrive UX innovation on Windows devices powered by Snapdragon
54mHarnessing On-Device AI to Transform the Smartphone Experience
14mAccelerating On-Device AI Development
38mBring AI experiences to all your Windows Devices
08mUnlocking AI at Scale for PyTorch Developers
17mOverview of Qualcomm Snapdragon Neural Processing Engine