Sustainable AI Processing at the Edge

TL;DR

This paper examines the environmental impacts of different edge computing accelerators for neural networks, comparing energy efficiency and sustainability of PIM, GPU, FPGA, and Racetrack memory solutions.

Contribution

It introduces a comprehensive analysis of the tradeoffs between various neural network acceleration technologies focusing on environmental sustainability and embodied energy.

Findings

Racetrack memory PIM can recover embodied energy in about 1 year.

Mobile GPUs are more sustainable at high activity ratios but have higher embodied energy.

PIM-enabled Racetrack memory offers significant energy recovery benefits.

Abstract

Edge computing is a popular target for accelerating machine learning algorithms supporting mobile devices without requiring the communication latencies to handle them in the cloud. Edge deployments of machine learning primarily consider traditional concerns such as SWaP constraints (Size, Weight, and Power) for their installations. However, such metrics are not entirely sufficient to consider environmental impacts from computing given the significant contributions from embodied energy and carbon. In this paper we explore the tradeoffs of convolutional neural network acceleration engines for both inference and on-line training. In particular, we explore the use of processing-in-memory (PIM) approaches, mobile GPU accelerators, and recently released FPGAs, and compare them with novel Racetrack memory PIM. Replacing PIM-enabled DDR3 with Racetrack memory PIM can recover its embodied energy as quickly as 1 year. For high activity ratios, mobile GPUs can be more sustainable but have higher embodied energy to overcome compared to PIM-enabled Racetrack memory.