An Energy-Efficient Digital ReRAM-Crossbar-Based CNN With Bitwise Parallelism
There is great attention to develop hardware accelerator with better energy efficiency, as well as throughput, than GPUs for convolutional neural network (CNN). The existing solutions have relatively limited parallelism as well as large power consumption (including leakage power). In this paper, we...
Saved in:
Main Authors: | , , , |
---|---|
Other Authors: | |
Format: | Article |
Language: | English |
Published: |
2017
|
Subjects: | |
Online Access: | https://hdl.handle.net/10356/85536 http://hdl.handle.net/10220/43795 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
id |
sg-ntu-dr.10356-85536 |
---|---|
record_format |
dspace |
spelling |
sg-ntu-dr.10356-855362020-03-07T13:57:27Z An Energy-Efficient Digital ReRAM-Crossbar-Based CNN With Bitwise Parallelism Ni, Leibin Liu, Zichuan Yu, Hao Joshi, Rajiv V. School of Electrical and Electronic Engineering Approximate computing Neural network hardware There is great attention to develop hardware accelerator with better energy efficiency, as well as throughput, than GPUs for convolutional neural network (CNN). The existing solutions have relatively limited parallelism as well as large power consumption (including leakage power). In this paper, we present a resistive random access memory (ReRAM)-accelerated CNN that can achieve significantly higher throughput and energy efficiency when the CNN is trained with binary constraints on both weights and activations, and is further mapped on a digital ReRAM-crossbar. We propose an optimized accelerator architecture tailored for bitwise convolution that features massive parallelism with high energy efficiency. Numerical experiment results show that the binary CNN accelerator on a digital ReRAM-crossbar achieves a peak throughput of 792 GOPS at the power consumption of 4.5 mW, which is 1.61 times faster and 296 times more energy-efficient than a high-end GPU. NRF (Natl Research Foundation, S’pore) MOE (Min. of Education, S’pore) Accepted version 2017-09-26T07:52:12Z 2019-12-06T16:05:32Z 2017-09-26T07:52:12Z 2019-12-06T16:05:32Z 2017 Journal Article Ni, L., Liu, Z., Yu, H., & Joshi, R. V. (2017). An Energy-Efficient Digital ReRAM-Crossbar-Based CNN With Bitwise Parallelism. IEEE Journal on Exploratory Solid-State Computational Devices and Circuits, 3, 37-46. https://hdl.handle.net/10356/85536 http://hdl.handle.net/10220/43795 10.1109/JXCDC.2017.2697910 en IEEE Journal on Exploratory Solid-State Computational Devices and Circuits © 2017 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. The published version is available at: [http://dx.doi.org/10.1109/JXCDC.2017.2697910]. 8 p. application/pdf |
institution |
Nanyang Technological University |
building |
NTU Library |
country |
Singapore |
collection |
DR-NTU |
language |
English |
topic |
Approximate computing Neural network hardware |
spellingShingle |
Approximate computing Neural network hardware Ni, Leibin Liu, Zichuan Yu, Hao Joshi, Rajiv V. An Energy-Efficient Digital ReRAM-Crossbar-Based CNN With Bitwise Parallelism |
description |
There is great attention to develop hardware accelerator with better energy efficiency, as well as throughput, than GPUs for convolutional neural network (CNN). The existing solutions have relatively limited parallelism as well as large power consumption (including leakage power). In this paper, we present a resistive random access memory (ReRAM)-accelerated CNN that can achieve significantly higher throughput and energy efficiency when the CNN is trained with binary constraints on both weights and activations, and is further mapped on a digital ReRAM-crossbar. We propose an optimized accelerator architecture tailored for bitwise convolution that features massive parallelism with high energy efficiency. Numerical experiment results show that the binary CNN accelerator on a digital ReRAM-crossbar achieves a peak throughput of 792 GOPS at the power consumption of 4.5 mW, which is 1.61 times faster and 296 times more energy-efficient than a high-end GPU. |
author2 |
School of Electrical and Electronic Engineering |
author_facet |
School of Electrical and Electronic Engineering Ni, Leibin Liu, Zichuan Yu, Hao Joshi, Rajiv V. |
format |
Article |
author |
Ni, Leibin Liu, Zichuan Yu, Hao Joshi, Rajiv V. |
author_sort |
Ni, Leibin |
title |
An Energy-Efficient Digital ReRAM-Crossbar-Based CNN With Bitwise Parallelism |
title_short |
An Energy-Efficient Digital ReRAM-Crossbar-Based CNN With Bitwise Parallelism |
title_full |
An Energy-Efficient Digital ReRAM-Crossbar-Based CNN With Bitwise Parallelism |
title_fullStr |
An Energy-Efficient Digital ReRAM-Crossbar-Based CNN With Bitwise Parallelism |
title_full_unstemmed |
An Energy-Efficient Digital ReRAM-Crossbar-Based CNN With Bitwise Parallelism |
title_sort |
energy-efficient digital reram-crossbar-based cnn with bitwise parallelism |
publishDate |
2017 |
url |
https://hdl.handle.net/10356/85536 http://hdl.handle.net/10220/43795 |
_version_ |
1681049866160046080 |