Advances in deep neural networks (DNNs) and the availability of massive real-world data have enabled superhuman levels of accuracy on many AI tasks and ushered the explosive growth of AI workloads across the spectrum of computing devices. However, their superior accuracy comes at a high computational cost, which necessitates approaches beyond traditional computing paradigms to improve their operational efficiency. Leveraging the application-level insight of error resilience, we demonstrate how approximate computing (AxC) can significantly boost the efficiency of AI platforms and play a pivotal role in the broader adoption of AI-based applications and services. To this end, we present RaPiD, a multi-tera operations per second (TOPS) AI hardware accelerator core (fabricated at 14-nm technology) that we built from the ground-up using AxC techniques across the stack including algorithms, architecture, programmability, and hardware. We highlight the workload-guided systematic explorations of AxC techniques for AI, including custom number representations, quantization/pruning methodologies, mixed-precision architecture design, instruction sets, and compiler technologies with quality programmability, employed in the RaPiD accelerator.

The combination of growth in compute capabilities and availability of large datasets has led to a re-birth of deep learning. Deep Neural Networks (DNNs) have become state-of-the-art in a variety of machine learning tasks spanning domains across vision, speech, and machine translation. Deep Learning (DL) achieves high accuracy in these tasks at the expense of 100s of ExaOps of computation; posing significant challenges to efficient large-scale deployment in both resource-constrained environments and data centers.

Starting with a brief review on 0.1-/spl mu/m (100 nm) CMOS status, this paper addresses the key challenges in further scaling of CMOS technology into the nanometer (sub-100 nm) regime in light of fundamental physical effects and practical considerations. Among the issues discussed are: lithography, power supply and threshold voltage, short-channel effect, gate oxide, high-field effects, dopant number fluctuations and interconnect delays. The last part of the paper discusses several alternative or unconventional device structures, including silicon-on-insulator (SOI), SiGe MOSFET's, low-temperature CMOS, and double-gate MOSFET's, which may lead to the outermost limits of silicon scaling.