A near-threshold memristive computing-in-memory engine for edge intelligence

Abstract Memristive computing-in-memory and near-threshold computing are two unconventional computing paradigms that can potentially enhance the energy efficiency and real-time performance of edge devices. However, their scalability faces challenges, primarily due to process variation. Here, we repo...

Full description

Saved in:
Bibliographic Details
Main Authors: Linfang Wang, Weizeng Li, Zhidao Zhou, Junjie An, Wang Ye, Zhi Li, Hanghang Gao, Hongyang Hu, Jing Liu, Xiaoming Chen, Ling Li, Qi Liu, Mingoo Seok, Chunmeng Dou, Ming Liu
Format: Article
Language:English
Published: Nature Portfolio 2025-07-01
Series:Nature Communications
Online Access:https://doi.org/10.1038/s41467-025-61025-4
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849769254827589632
author Linfang Wang
Weizeng Li
Zhidao Zhou
Junjie An
Wang Ye
Zhi Li
Hanghang Gao
Hongyang Hu
Jing Liu
Xiaoming Chen
Ling Li
Qi Liu
Mingoo Seok
Chunmeng Dou
Ming Liu
author_facet Linfang Wang
Weizeng Li
Zhidao Zhou
Junjie An
Wang Ye
Zhi Li
Hanghang Gao
Hongyang Hu
Jing Liu
Xiaoming Chen
Ling Li
Qi Liu
Mingoo Seok
Chunmeng Dou
Ming Liu
author_sort Linfang Wang
collection DOAJ
description Abstract Memristive computing-in-memory and near-threshold computing are two unconventional computing paradigms that can potentially enhance the energy efficiency and real-time performance of edge devices. However, their scalability faces challenges, primarily due to process variation. Here, we report a 1-Mb, 16-macro near-threshold memristive computing-in-memory engine. The two-transistor-one-resistor cells provide strong cell current modulation capability with more than 120-times amplified resistance ratio. To mitigate variation issues, we compensate for transistor mismatches by leveraging the intrinsic variations in memristors. Additionally, we propose a charge stacking technique between multiple analog-to-digital converters to perform analog weight-and-combine operations with small energy and area overhead. Moreover, we introduce an inter-macro hybrid control scheme to reduce the task-level inference power. The fabricated chip can perform highly parallel analog computing over 256 input channels with a 2.4% relative standard deviation. It achieves a throughput up to 10.49 tera-operations per second and an energy efficiency up to 88.51 tera-operations per second per watt.
format Article
id doaj-art-0aeb71f5872e462b8814648b36e6cfb3
institution DOAJ
issn 2041-1723
language English
publishDate 2025-07-01
publisher Nature Portfolio
record_format Article
series Nature Communications
spelling doaj-art-0aeb71f5872e462b8814648b36e6cfb32025-08-20T03:03:28ZengNature PortfolioNature Communications2041-17232025-07-0116111010.1038/s41467-025-61025-4A near-threshold memristive computing-in-memory engine for edge intelligenceLinfang Wang0Weizeng Li1Zhidao Zhou2Junjie An3Wang Ye4Zhi Li5Hanghang Gao6Hongyang Hu7Jing Liu8Xiaoming Chen9Ling Li10Qi Liu11Mingoo Seok12Chunmeng Dou13Ming Liu14State Key Lab of Fabrication Technologies for Integrated Circuits, Institute of Microelectronics of the Chinese Academy of SciencesState Key Lab of Fabrication Technologies for Integrated Circuits, Institute of Microelectronics of the Chinese Academy of SciencesState Key Lab of Fabrication Technologies for Integrated Circuits, Institute of Microelectronics of the Chinese Academy of SciencesState Key Lab of Fabrication Technologies for Integrated Circuits, Institute of Microelectronics of the Chinese Academy of SciencesState Key Lab of Fabrication Technologies for Integrated Circuits, Institute of Microelectronics of the Chinese Academy of SciencesState Key Lab of Fabrication Technologies for Integrated Circuits, Institute of Microelectronics of the Chinese Academy of SciencesState Key Lab of Fabrication Technologies for Integrated Circuits, Institute of Microelectronics of the Chinese Academy of SciencesState Key Lab of Fabrication Technologies for Integrated Circuits, Institute of Microelectronics of the Chinese Academy of SciencesState Key Lab of Fabrication Technologies for Integrated Circuits, Institute of Microelectronics of the Chinese Academy of SciencesInstitute of Computing Technology of the Chinese Academy of SciencesInstitute of Software of the Chinese Academy of SciencesState Key Lab of Fabrication Technologies for Integrated Circuits, Institute of Microelectronics of the Chinese Academy of SciencesDepartment of Electrical Engineering, Columbia UniversityState Key Lab of Fabrication Technologies for Integrated Circuits, Institute of Microelectronics of the Chinese Academy of SciencesState Key Lab of Fabrication Technologies for Integrated Circuits, Institute of Microelectronics of the Chinese Academy of SciencesAbstract Memristive computing-in-memory and near-threshold computing are two unconventional computing paradigms that can potentially enhance the energy efficiency and real-time performance of edge devices. However, their scalability faces challenges, primarily due to process variation. Here, we report a 1-Mb, 16-macro near-threshold memristive computing-in-memory engine. The two-transistor-one-resistor cells provide strong cell current modulation capability with more than 120-times amplified resistance ratio. To mitigate variation issues, we compensate for transistor mismatches by leveraging the intrinsic variations in memristors. Additionally, we propose a charge stacking technique between multiple analog-to-digital converters to perform analog weight-and-combine operations with small energy and area overhead. Moreover, we introduce an inter-macro hybrid control scheme to reduce the task-level inference power. The fabricated chip can perform highly parallel analog computing over 256 input channels with a 2.4% relative standard deviation. It achieves a throughput up to 10.49 tera-operations per second and an energy efficiency up to 88.51 tera-operations per second per watt.https://doi.org/10.1038/s41467-025-61025-4
spellingShingle Linfang Wang
Weizeng Li
Zhidao Zhou
Junjie An
Wang Ye
Zhi Li
Hanghang Gao
Hongyang Hu
Jing Liu
Xiaoming Chen
Ling Li
Qi Liu
Mingoo Seok
Chunmeng Dou
Ming Liu
A near-threshold memristive computing-in-memory engine for edge intelligence
Nature Communications
title A near-threshold memristive computing-in-memory engine for edge intelligence
title_full A near-threshold memristive computing-in-memory engine for edge intelligence
title_fullStr A near-threshold memristive computing-in-memory engine for edge intelligence
title_full_unstemmed A near-threshold memristive computing-in-memory engine for edge intelligence
title_short A near-threshold memristive computing-in-memory engine for edge intelligence
title_sort near threshold memristive computing in memory engine for edge intelligence
url https://doi.org/10.1038/s41467-025-61025-4
work_keys_str_mv AT linfangwang anearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT weizengli anearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT zhidaozhou anearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT junjiean anearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT wangye anearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT zhili anearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT hanghanggao anearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT hongyanghu anearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT jingliu anearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT xiaomingchen anearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT lingli anearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT qiliu anearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT mingooseok anearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT chunmengdou anearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT mingliu anearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT linfangwang nearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT weizengli nearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT zhidaozhou nearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT junjiean nearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT wangye nearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT zhili nearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT hanghanggao nearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT hongyanghu nearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT jingliu nearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT xiaomingchen nearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT lingli nearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT qiliu nearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT mingooseok nearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT chunmengdou nearthresholdmemristivecomputinginmemoryengineforedgeintelligence
AT mingliu nearthresholdmemristivecomputinginmemoryengineforedgeintelligence