The custom silicon is a coprocessor designed by Microsoft that will natively and flexibly implement DNNs, according to the company.
Instead of relying on FPGAs to provide cheaper and efficient AI execution and to form deep neural networks (DNNs), Microsoft has announced that it will incorporate a custom silicon into its second-generation of holographic processing unit (HPU).
The custom silicon is a "coprocessor" designed but not manufactured by Microsoft that will "natively and flexibly implement DNNs," Marc Pollefeys, Microsoft's director of science for HoloLens, wrote in blog post.
The HoloLen's HPU is a custom multiprocessor that is responsible for processing the information coming from all of the on-board sensors, including Microsoft’s custom time-of-flight depth sensor, head-tracking cameras, the inertial measurement unit (IMU) and the infrared camera.
For the second version of the HPU, Pollefeys said they will add the AI coprocessor, which supports a wide variety of fully programmable layer types.
"The AI coprocessor is designed to work in the next version of HoloLens, running continuously, off the HoloLens battery. This is just one example of the new capabilities we are developing for HoloLens," Pollefeys said.