Microsoft has revealed that it is designing its own custom chip for the upcoming HoloLens 2. According to a blog post on Microsoft Research Blog, the company pointed out that the new custom chip will handle AI functions on the HoloLens 2 where it will analyze visual data on the device and save time by not uploading the data that is not required.
The chip will act as a ‘coprocessor’ for the sensors on the device and the company will not manufacture it in-house. This level of ‘on-device’ processing will help with performance on the second generation of the mixed reality headset. The interesting thing to note here is that Microsoft is not the only company that is following the trend of designing its own processor.
Marc Pollefeys, the Director of Science for HoloLens department at Microsoft pointed out that this ‘coprocessor’ will ensure that the company is able to incorporate Deep Neural Networks (DNNs) while ensuring that HoloLens 2 is more user friendly and more secure. The device will not require constant internet connection and user data will not leave the device. The team working on the new chip has dubbed it as HPU 2.0 (Holographic Processing Unit).
Harry Shum, the executive Vice President of AI and Research Group at Microsoft announced the company’s decision to design custom chip at CVPR 2017. The company has not revealed the date when HPU 2.0 will be ready or when the HoloLens 2.0 will hit the market but it pointed out that the processor is ;currently under development.’