16.4 An 89TOPS/W and 16.3TOPS/mm<sup>2</sup> All-Digital SRAM-Based Full-Precision Compute-In Memory Macro in 22nm for Machine-Learning Edge Applications

Yu-Der Chih, Po-Hao Lee, Hidehiro Fujiwara, Yi-Chun Shih, Chia-Fu Lee, Rawan Naous, Yulin Chen, Chieh-Pu Lo, C.W. Lu, H. Mori,

2021 · 335 citations · 6 references

Concepts

TL;DR

Artificial intelligence and machine learning are increasingly deployed on edge devices, but conventional CPUs, GPUs, FPGAs, and ASICs cannot satisfy the energy constraints of future edge applications because most of the power is consumed by memory accesses. The study aims to enhance energy efficiency by investigating compute‑in‑memory (CIM) architectures. CIM research focuses on an analog approach that offers high energy efficiency, yet suffers from low signal‑to‑noise ratio, limiting its accuracy for high‑precision applications.

Abstract

From the cloud to edge devices, artificial intelligence (AI) and machine learning (ML) are widely used in many cognitive tasks, such as image classification and speech recognition. In recent years, research on hardware accelerators for AI edge devices has received more attention, mainly due to the advantages of AI at the edge: including privacy, low latency, and more reliable and effective use of network bandwidth. However, traditional computing architectures (such as CPUs, GPUs, FPGAs, and even existing AI accelerator ASICs) cannot meet the future needs of energy-constrained AI edge applications. This is because ML computing is data-centric, most of the energy in these architectures is consumed by memory accesses. In order to improve energy efficiency, both academia and industry are exploring a new computing architecture, namely compute in memory (CIM). CIM research is focused on a more analog approach with high-energy efficiency; however, lack of accuracy, due to a low SNR, is the main disadvantage; therefore, an analog approach may not be suitable for some applications that require high accuracy.

References

6