VLSI scaling has led to the advent of massively parallel computing architectures as a way to circumvent the limitations of conventional approaches to improving performance and power. This can be seen in the architecture of general-purpose processors and in the advent of general-purpose graphics processing units. Standard architectures face algorithmic limitations to parallelism that limits scaling. Neuromorphic computing provides a promising alternative as a natively parallel computing paradigm for implementing real-time processing for vision and speech. We present our recent experiences with the design of TrueNorth, a 4096-core massively parallel neuromorphic computing substrate that was developed through a deep collaboration with IBM research.

Rajit Manohar is Professor of Electrical and Computer Engineering and a Stephen H. Weiss Presidential Fellow at Cornell. He received his B.S. (1994), M.S. (1995), and Ph.D. (1998) from Caltech. He has been on the Cornell faculty since 1998 and the Cornell Tech faculty since 2012, where his group conducts research on self-timed systems. He is the recipient of an NSF CAREER award, seven best paper awards, seven teaching awards, and was named to MIT technology review's top 35 young innovators under 35 for contributions to low power microprocessor design.