Qualcomm products mentioned within this post are offered by
Qualcomm Technologies, Inc. and/or its subsidiaries.
Today at Microsoft Build 2022, Qualcomm Technologies is announcing that we are opening up our industry-leading Qualcomm AI Engine for Windows developers with the Qualcomm Neural Processing SDK for Windows. In addition, Qualcomm Technologies and Microsoft are announcing Project Volterra, a new Windows Developer Kit powered by the Snapdragon Compute Platform.
The Qualcomm Neural Processing SDK for Windows gives developers the tools they need to hardware-accelerate ML model inference in their Windows apps when running on Snapdragon. This was previously only possible for Android using the regular SDK. This rich SDK includes pipeline tools to convert and optimize ML models from popular ML frameworks for Qualcomm Hexagon DSP processors, and a runtime side (API) to load and execute models on Hexagon. It also supports popular frameworks like TensorFlow and ONNX. Using it frees up the GPU and CPU, while the Hexagon DSP performs the model inference computations in parallel, giving you the best possible power efficiency and performance.
Project Volterra will be built on the Snapdragon compute platform, which incorporates our Hexagon processor, the core of our Qualcomm AI Engine. Developers will be able to build Windows applications that execute on-device AI-accelerated workloads with extreme performance and efficiency.
Developers will use Project Volterra Kit to build, test, and debug ARM64, ARM64EC, and AI applications accelerated on the Hexagon processor in Windows. Best of all, you can use the dev kit alongside all of your favorite productivity tools, including Windows Terminal, WSL, Microsoft Office, and Teams, and of course, the much-loved Visual Studio and Visual Studio Code IDEs.
Project Volterra will complement the Snapdragon Developer Kit for Windows-based on our Snapdragon 7c Compute Platform (aka the ECS LIVA Mini Box QC710 Desktop, also available from Microsoft), which you may already use. Snapdragon compute platforms allow you to do more for Windows on-device AI. Build features like facial detection, background blur, audio blur, and transcription for use cases that span productivity, collaboration, and security from anywhere, delivered with extreme efficiency.
For example, Snapdragon can perform gaze correction using around 300mW versus the approximate 15 watts of power required by a competitor’s x86-based PC thanks to the Hexagon processor. This represents 50 times less power consumption and significantly increases battery life while not compromising CPU or GPU performance.
Also check out our latest AI-powered noise suppression on the new Lenovo ThinkPad X13s, which can effectively remove background noise while someone speaks. You can see how the technology can isolate a speaker’s voice from different background noises ranging from crinkling snack bags to drums.
Even with these high-performance features, compute cycles are reserved so that developers can optimize business logic and rendering performance, while the Hexagon processor handles the AI load in parallel. The net result: a platform that can power the ultimate mobile PC, backed by a dev kit and the Qualcomm Neural Processing SDK for Windows on-device AI that we think developers will love.
Currently, 200 enterprises are now testing or deploying Windows on Snapdragon PCs, and the Lenovo ThinkPad X13s will feature the premium Snapdragon Compute Platform.
If you are attending Build 2022, check out our video on-demand sessions:
- Tapping into the Qualcomm AI Engine – On-device AI driving improved and new Windows experiences
- Discuss Performance and Efficiency with On-Device AI and Windows Experiences
For additional information, be sure to check out the following resources:
- Windows on Snapdragon Landing Page
- WoS Developer Resources
- Snapdragon Developer Kit
- WoS Learning Resources
- Qualcomm Neural Processing SDK for Windows
Snapdragon, Qualcomm Neural Processing SDK, Qualcomm Hexagon, and Qualcomm AI Engine are products of Qualcomm Technologies, Inc. and/or its subsidiaries.