Variability-Aware Design for Energy Efficient Computational Artificial Intelligence Platform

TL;DR

This paper discusses the importance of variability-aware design strategies to improve energy efficiency in AI platforms, especially for portable devices affected by manufacturing, environmental, and aging variability.

Contribution

It introduces a variability characterization platform to address power consumption variability in energy-efficient AI hardware design.

Findings

Variability significantly impacts power consumption in portable AI devices.

A new platform for variability characterization can help optimize energy efficiency.

Variability effects become more pronounced at smaller technology nodes.

Abstract

Portable computing devices, which include tablets, smart phones and various types of wearable sensors, experienced a rapid development in recent years. One of the most critical limitations for these devices is the power consumption as they use batteries as the power supply. However, the bottleneck of the power saving schemes in both hardware design and software algorithm is the huge variability in power consumption. The variability is caused by a myriad of factors, including the manufacturing process, the ambient environment (temperature, humidity), the aging effects and etc. As the technology node scaled down to 28nm and even lower, the variability becomes more severe. As a result, a platform for variability characterization seems to be very necessary and helpful.