Microsoft, in collaboration with Qualcomm, has prepared Project Volterra. It is a mini-computer similar to the Mac mini, on which developers will be able to file a new generation of applications. Those that take advantage of hardware acceleration of AI calculations.
The company will bring an interface to Windows 11 that will allow the application to access neural coprocessors (NPUs), which will process machine-based tasks – typically image and audio processing – faster than the CPU or GPU.
Qualcomm integrates the NPU Hexagon into its chips, but Microsoft expects to find its counterpart in most or perhaps all future computers. Therefore, Intel and AMD will include them one day. In the mobile world, NPUs have been established for years, we can also find them in SoC from Samsung (Exynos), Huawei (Kirin) or MediaTek (Dimensity), where they mainly make sure that the photos are sharper, noiseless and generally nicer. In this respect, Apple is the furthest to integrate the NPU called Neural Engine into mobile and desktop chips and counts on it in both iOS and macOS.
By introducing a new AI interface, Microsoft will make it easier for developers, as they would not have to support different accelerators from different vendors, but will use an equivalent universal API such as DirectX.
At the moment, however, we do not yet have more precise information about that interface or Project Volterra. We may learn more about some of the accompanying lectures at the Build conference, or a few weeks later when Microsoft releases a new version of development tools.