Abundant-data applications of the 21st century will arise from social networks, e-commerce transactions, genome sequences, and trillions of interconnected sensors which could create a data deluge. Future technologies must be able to process and classify such data in real-time, in an energy-efficient manner, a challenge which will expose gross inefficiencies in traditional computing architectures. This short course will introduce the Nano-Engineered Computing Systems Technology (N3XT) approach [1] within the context of the Stanford SystemX Alliance [2], with particular focus on heterogeneous integration of low-power nanoelectronics. Our groups of researchers within SystemX are pursuing the 3D monolithic integration of logic, memory and thermal management, as well as energy-efficient and energy-harvesting design opportunities. This approach capitalizes on several recent nanotechnology breakthroughs, such as high-performance and energy-efficient 1D and 2D transistors, non-volatile memory such as RRAM, 3D fine-grained interconnects, and embedded cooling technologies based on conduction and convection breakthroughs. The benefits of monolithically co-integrating logic and memory into 3D systems can be up to 1000-fold in terms of energy-delay-product (EDP) benefits, especially for future abundant-data applications.
[1] M. Aly et al., Computer 48, 24-33 (2015)
[2] https://systemx.stanford.edu