24.5 A Twin-8T SRAM Computation-In-Memory Macro for Multiple-Bit CNN-Based Machine Learning

Xin Si, Jia Jing Chen, Yung Ning Tu, Wei Hsing Huang, Jing Hong Wang, Yen Cheng Chiu, Wei Chen Wei, Ssu Yen Wu, Xiaoyu Sun, Rui Liu, Shimeng Yu, Ren Shuo Liu, Chih Cheng Hsieh, Kea Tiong Tang, Qiang Li, Meng Fan Chang

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Citation (Scopus)

Abstract

Computation-in-memory (CIM) is a promising avenue to improve the energy efficiency of multiply-and-accumulate (MAC) operations in AI chips. Multi-bit CNNs are required for high-inference accuracy in many applications [1-5]. There are challenges and tradeoffs for SRAM-based CIM: (1) tradeoffs between signal margin, cell stability and area overhead; (2) the high-weighted bit process variation dominates the end-result error rate; (3) trade-off between input bandwidth, speed and area. Previous SRAM CIM macros were limited to binary MAC operations for fully connected networks [1], or they used CIM for multiplication [2] or weight-combination operations [3] with additional large-area near-memory computing (NMC) logic for summation or MAC operations.

Original languageEnglish (US)
Title of host publication2019 IEEE International Solid-State Circuits Conference, ISSCC 2019
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages396-398
Number of pages3
ISBN (Electronic)9781538685310
DOIs
StatePublished - Mar 6 2019
Event2019 IEEE International Solid-State Circuits Conference, ISSCC 2019 - San Francisco, United States
Duration: Feb 17 2019Feb 21 2019

Publication series

NameDigest of Technical Papers - IEEE International Solid-State Circuits Conference
Volume2019-February
ISSN (Print)0193-6530

Conference

Conference2019 IEEE International Solid-State Circuits Conference, ISSCC 2019
CountryUnited States
CitySan Francisco
Period2/17/192/21/19

Fingerprint

Static random access storage
Macros
Learning systems
Data storage equipment
Energy efficiency
Bandwidth

ASJC Scopus subject areas

  • Electronic, Optical and Magnetic Materials
  • Electrical and Electronic Engineering

Cite this

Si, X., Chen, J. J., Tu, Y. N., Huang, W. H., Wang, J. H., Chiu, Y. C., ... Chang, M. F. (2019). 24.5 A Twin-8T SRAM Computation-In-Memory Macro for Multiple-Bit CNN-Based Machine Learning. In 2019 IEEE International Solid-State Circuits Conference, ISSCC 2019 (pp. 396-398). [8662392] (Digest of Technical Papers - IEEE International Solid-State Circuits Conference; Vol. 2019-February). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ISSCC.2019.8662392

24.5 A Twin-8T SRAM Computation-In-Memory Macro for Multiple-Bit CNN-Based Machine Learning. / Si, Xin; Chen, Jia Jing; Tu, Yung Ning; Huang, Wei Hsing; Wang, Jing Hong; Chiu, Yen Cheng; Wei, Wei Chen; Wu, Ssu Yen; Sun, Xiaoyu; Liu, Rui; Yu, Shimeng; Liu, Ren Shuo; Hsieh, Chih Cheng; Tang, Kea Tiong; Li, Qiang; Chang, Meng Fan.

2019 IEEE International Solid-State Circuits Conference, ISSCC 2019. Institute of Electrical and Electronics Engineers Inc., 2019. p. 396-398 8662392 (Digest of Technical Papers - IEEE International Solid-State Circuits Conference; Vol. 2019-February).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Si, X, Chen, JJ, Tu, YN, Huang, WH, Wang, JH, Chiu, YC, Wei, WC, Wu, SY, Sun, X, Liu, R, Yu, S, Liu, RS, Hsieh, CC, Tang, KT, Li, Q & Chang, MF 2019, 24.5 A Twin-8T SRAM Computation-In-Memory Macro for Multiple-Bit CNN-Based Machine Learning. in 2019 IEEE International Solid-State Circuits Conference, ISSCC 2019., 8662392, Digest of Technical Papers - IEEE International Solid-State Circuits Conference, vol. 2019-February, Institute of Electrical and Electronics Engineers Inc., pp. 396-398, 2019 IEEE International Solid-State Circuits Conference, ISSCC 2019, San Francisco, United States, 2/17/19. https://doi.org/10.1109/ISSCC.2019.8662392
Si X, Chen JJ, Tu YN, Huang WH, Wang JH, Chiu YC et al. 24.5 A Twin-8T SRAM Computation-In-Memory Macro for Multiple-Bit CNN-Based Machine Learning. In 2019 IEEE International Solid-State Circuits Conference, ISSCC 2019. Institute of Electrical and Electronics Engineers Inc. 2019. p. 396-398. 8662392. (Digest of Technical Papers - IEEE International Solid-State Circuits Conference). https://doi.org/10.1109/ISSCC.2019.8662392
Si, Xin ; Chen, Jia Jing ; Tu, Yung Ning ; Huang, Wei Hsing ; Wang, Jing Hong ; Chiu, Yen Cheng ; Wei, Wei Chen ; Wu, Ssu Yen ; Sun, Xiaoyu ; Liu, Rui ; Yu, Shimeng ; Liu, Ren Shuo ; Hsieh, Chih Cheng ; Tang, Kea Tiong ; Li, Qiang ; Chang, Meng Fan. / 24.5 A Twin-8T SRAM Computation-In-Memory Macro for Multiple-Bit CNN-Based Machine Learning. 2019 IEEE International Solid-State Circuits Conference, ISSCC 2019. Institute of Electrical and Electronics Engineers Inc., 2019. pp. 396-398 (Digest of Technical Papers - IEEE International Solid-State Circuits Conference).
@inproceedings{315e368fdd1d4d6baa4e0f4f79ba6d1e,
title = "24.5 A Twin-8T SRAM Computation-In-Memory Macro for Multiple-Bit CNN-Based Machine Learning",
abstract = "Computation-in-memory (CIM) is a promising avenue to improve the energy efficiency of multiply-and-accumulate (MAC) operations in AI chips. Multi-bit CNNs are required for high-inference accuracy in many applications [1-5]. There are challenges and tradeoffs for SRAM-based CIM: (1) tradeoffs between signal margin, cell stability and area overhead; (2) the high-weighted bit process variation dominates the end-result error rate; (3) trade-off between input bandwidth, speed and area. Previous SRAM CIM macros were limited to binary MAC operations for fully connected networks [1], or they used CIM for multiplication [2] or weight-combination operations [3] with additional large-area near-memory computing (NMC) logic for summation or MAC operations.",
author = "Xin Si and Chen, {Jia Jing} and Tu, {Yung Ning} and Huang, {Wei Hsing} and Wang, {Jing Hong} and Chiu, {Yen Cheng} and Wei, {Wei Chen} and Wu, {Ssu Yen} and Xiaoyu Sun and Rui Liu and Shimeng Yu and Liu, {Ren Shuo} and Hsieh, {Chih Cheng} and Tang, {Kea Tiong} and Qiang Li and Chang, {Meng Fan}",
year = "2019",
month = "3",
day = "6",
doi = "10.1109/ISSCC.2019.8662392",
language = "English (US)",
series = "Digest of Technical Papers - IEEE International Solid-State Circuits Conference",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
pages = "396--398",
booktitle = "2019 IEEE International Solid-State Circuits Conference, ISSCC 2019",

}

TY - GEN

T1 - 24.5 A Twin-8T SRAM Computation-In-Memory Macro for Multiple-Bit CNN-Based Machine Learning

AU - Si, Xin

AU - Chen, Jia Jing

AU - Tu, Yung Ning

AU - Huang, Wei Hsing

AU - Wang, Jing Hong

AU - Chiu, Yen Cheng

AU - Wei, Wei Chen

AU - Wu, Ssu Yen

AU - Sun, Xiaoyu

AU - Liu, Rui

AU - Yu, Shimeng

AU - Liu, Ren Shuo

AU - Hsieh, Chih Cheng

AU - Tang, Kea Tiong

AU - Li, Qiang

AU - Chang, Meng Fan

PY - 2019/3/6

Y1 - 2019/3/6

N2 - Computation-in-memory (CIM) is a promising avenue to improve the energy efficiency of multiply-and-accumulate (MAC) operations in AI chips. Multi-bit CNNs are required for high-inference accuracy in many applications [1-5]. There are challenges and tradeoffs for SRAM-based CIM: (1) tradeoffs between signal margin, cell stability and area overhead; (2) the high-weighted bit process variation dominates the end-result error rate; (3) trade-off between input bandwidth, speed and area. Previous SRAM CIM macros were limited to binary MAC operations for fully connected networks [1], or they used CIM for multiplication [2] or weight-combination operations [3] with additional large-area near-memory computing (NMC) logic for summation or MAC operations.

AB - Computation-in-memory (CIM) is a promising avenue to improve the energy efficiency of multiply-and-accumulate (MAC) operations in AI chips. Multi-bit CNNs are required for high-inference accuracy in many applications [1-5]. There are challenges and tradeoffs for SRAM-based CIM: (1) tradeoffs between signal margin, cell stability and area overhead; (2) the high-weighted bit process variation dominates the end-result error rate; (3) trade-off between input bandwidth, speed and area. Previous SRAM CIM macros were limited to binary MAC operations for fully connected networks [1], or they used CIM for multiplication [2] or weight-combination operations [3] with additional large-area near-memory computing (NMC) logic for summation or MAC operations.

UR - http://www.scopus.com/inward/record.url?scp=85063542939&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85063542939&partnerID=8YFLogxK

U2 - 10.1109/ISSCC.2019.8662392

DO - 10.1109/ISSCC.2019.8662392

M3 - Conference contribution

AN - SCOPUS:85063542939

T3 - Digest of Technical Papers - IEEE International Solid-State Circuits Conference

SP - 396

EP - 398

BT - 2019 IEEE International Solid-State Circuits Conference, ISSCC 2019

PB - Institute of Electrical and Electronics Engineers Inc.

ER -