A Comparative Analysis of Time-Domain and Digital-Domain Hardware Accelerators for Neural Networks

Hamza Al Maharmeh, Nabil J. Sarhan, Chung Chih Hung, Mohammed Ismail, Mohammad Alhawari

研究成果: Conference article同行評審

2 引文 斯高帕斯(Scopus)

摘要

This paper presents a comprehensive analysis of hardware accelerators for neural networks in both the digital and time domains, where the latter includes spatially unrolled (SU) and recursive (REC) architectures. All accelerators are implemented and synthesized in a 65nm CMOS technology. An identical neural network model is implemented in the digital and time domain for comparative purposes in terms of throughput, power consumption, area, and energy efficiency. Post-synthesis results show that SU achieves the highest energy efficiency of 145 TOp/s/W with a throughput of 4 GOp/s. The digital core is the fastest among other cores, whereas REC is the slowest but is the most area-efficient, occupying 0.114 mm2. SU is more suited for applications with stringent power constraints and average performance, while REC is better suited for applications where the area is the most important requirement and the throughput is less significant. In contrast, the digital core is preferable for large neural networks and critical applications that require high performance.

原文English
期刊Proceedings - IEEE International Symposium on Circuits and Systems
2021-January
DOIs
出版狀態Published - 2021
事件53rd IEEE International Symposium on Circuits and Systems, ISCAS 2021 - Daegu, Korea, Republic of
持續時間: 22 5月 202128 5月 2021

指紋

深入研究「A Comparative Analysis of Time-Domain and Digital-Domain Hardware Accelerators for Neural Networks」主題。共同形成了獨特的指紋。

引用此