Eyeriss: An Energy-Efficient Reconfigurable Accelerator for Deep Convolutional Neural Networks
The DOI matches its Crossref record; 38 of the 42 references it cites were found in a registry or an index.
Abstract
Eyeriss is an accelerator for state-of-the-art deep convolutional neural networks (CNNs). It optimizes for the energy efficiency of the entire system, including the accelerator chip and off-chip DRAM, for various CNN shapes by reconfiguring the architecture. CNNs are widely used in modern AI systems but also bring challenges on throughput and energy efficiency to the underlying hardware. This is because its computation requires a large amount of data, creating significant data movement from on-chip and off-chip that is more energy-consuming than computation. Minimizing data movement energy cost for any CNN shape, therefore, is the key to high throughput and energy efficiency. Eyeriss achieves these goals by using a proposed processing dataflow, called row stationary (RS), on a spatial architecture with 168 processing elements. RS dataflow reconfigures the computation mapping of a given shape, which optimizes energy efficiency by maximally reusing data locally to reduce expensive data movement, such as DRAM accesses. Compression and data gating are also applied to further improve energy efficiency. Eyeriss processes the convolutional layers at 35 frames/s and 0.0029 DRAM access/multiply and accumulation (MAC) for AlexNet at 278 mW (batch size N = 4), and 0.7 frames/s and 0.0035 DRAM access/MAC for VGG-16 at 236 mW (N = 3).
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“such scaling will enable NeuRRAM to deliver higher energy and area efficiency than today’s state-of-the-art edge inference accelerators58–61.”
From A compute-in-memory chip based on resistive random-access memory (2022) · cites this paper as [58]
In the citing paper
“Some studies proposed to implement the argmax() function via hardware128–148, which could be beneficial to reduce the total transistor count and power consumption while at the same time increasing the throughput.”
From Hardware implementation of memristor-based artificial neural networks (2024) · cites this paper as [131]
In the citing paper
“Based on the analysis of the research, CoNNA is able to speed up at least 3.08 times and up to 14.1 times faster than the MIT Eyeriss Accelerator [20].”
From Embedded Intelligence on FPGA: Survey, Applications and Challenges (2021) · cites this paper as [20]
In the citing paper
“massively parallel processing architectures with distributed memory architectures, such as graphic processing units (GPUs) but also increasingly specialized chips for NN inference (e.g., [17]), have shown to be very energy efficient and capable of impressive computational throughput when fully utilized by concurrent algorithms [18].”
From An Introduction to Deep Learning for the Physical Layer (2017) · cites this paper as [17]
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 1×
“Modern CNNs achieve their superior accuracy by building a very deep hierarchy of layers [2]–[5], [7], which transform the input image data into highly abstract representations called feature maps (fmaps).”
BackgroundVery Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148 · cited 7×
“We benchmark the chip performance using two publicly available and widely used CNNs: AlexNet [2] and VGG-16 [3].”
MethodInfluentialImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155
LeCun, Y., Bengio, Y., & Hinton, G. (2015) · Nature · doi:10.1038/nature14539 · Cited by 84,126
Gradient-based learning applied to document recognition
Lecun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998) · Proceedings of the IEEE · doi:10.1109/5.726791 · Cited by 58,805 · cited 1×
“[25] demonstrate a CNN processor and only report the theoretical peak throughput along with the power measured on a CNN for the MNIST data set [26], which has storage and computation requirements that are orders of magnitude lower than the state-of-the-art CNNs.”
MethodGoing deeper with convolutions
Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., & Rabinovich, A. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298594 · Cited by 48,562
ImageNet Large Scale Visual Recognition Challenge
Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al. (2015) · International Journal of Computer Vision · doi:10.1007/s11263-015-0816-y · Cited by 39,380 · cited 2×
“We have demonstrated an 1000-class image classification task [27] using this system, and a live demo can be found in [29].”
BackgroundMethodRich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation
Girshick, R., Donahue, J., Darrell, T., & Malik, J. (2014) · 2014 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2014.81 · Cited by 32,684
Mastering the game of Go with deep neural networks and tree search
Silver, D., Huang, A., Maddison, C. J., Guez, A., Sifre, L., van den Driessche, G., Schrittwieser, J., Antonoglou, I., Panneershelvam, V., Lanctot, M., et al. (2016) · Nature · doi:10.1038/nature16961 · Cited by 15,612 · cited 1×
“D EEP learning using convolutional neural networks (CNNs) [1] has achieved unprecedented accuracy on many modern AI applications [2]–[9].”
Background(2010) · International Conference on Machine Learning · Cited by 13,995
Research works printed but not linked: 2
- han, 2015, Learning both weights and connections for efficient neural network, Proc Adv Neural Inf Process Syst, 28, 1135
- chen, 2016, An 1000-Class Image Classification Task Performed by the Eyeriss-Integrated Deep Learning System
Cited by
Items 801–900 of 1,001
Li, A., Mo, H., Zhu, W., Qiang, L., Yin, S., Wei, S., & Liu, L. (2022) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2022.3146202 · Cited by 24
NAND-SPIN-based processing-in-MRAM architecture for convolutional neural network acceleration
Zhao, Y., Yang, J., Li, B., Cheng, X., Ye, X., Wang, X., Jia, X., Wang, Z., Zhang, Y., & Zhao, W. S. (2023) · Science China Information Sciences · doi:10.1007/s11432-021-3472-9 · Cited by 24
Enhancing Collective Communication in MCM Accelerators for Deep Learning Training
Laskar, S., Majhi, P., Kim, S., Mahmud, F., Muzahid, A., & Kim, E. J. (2024) · IEEE International Symposium on High-Performance Computer Architecture (HPCA) · doi:10.1109/hpca57654.2024.00069 · Cited by 32
“Simba [62] and Zimmer et al. [89] both present a 36-chiplet system in a 6 × 6 mesh, while Tangram [17] has a 16 × 16 chip-engine, where each engine similar to the Eyeriss [11] accelerator.”
MethodMao, R., Tang, L., Yuan, X., Ye, L., & Zhou, J. (2024) · IEEE International Symposium on High-Performance Computer Architecture (HPCA) · doi:10.1109/hpca57654.2024.00023 · Cited by 33
Zhang, J., Li, X., Xiao, P., Wei, Z., & Hong, Q. (2024) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2024.3391281 · Cited by 23
“Similarly, Chen et al. [18] also proposed a CNN accelerator called Eyeriss, which can minimize data access and support the high throughput CNN inference.”
MethodA Case of On-Chip Memory Subsystem Design for Low-Power CNN Accelerators
Wang, Y., Li, H., & Li, X. (2017) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2017.2778060 · Cited by 25
Lossy Compression for Embedded Computer Vision Systems
Guo, L., Zhou, D., Zhou, J., Kimura, S., & Goto, S. (2018) · IEEE Access · doi:10.1109/access.2018.2852809 · Cited by 22
Reducing MAC operation in convolutional neural network with sign prediction
Chang, J., Choi, Y., Lee, T., & Cho, J. (2018) · International Conference on Information and Communication Technology Convergence (ICTC) · doi:10.1109/ictc.2018.8539530 · Cited by 27
Kung, H. T., McDanel, B., & Zhang, S. Q. (2018) · SiPS ... design and implementation - IEEE Workshop on Signal Processing Systems · doi:10.1109/sips.2018.8598454 · Cited by 26
How to Measure Energy Consumption in Machine Learning Algorithms
García-Martín, E., Lavesson, N., Grahn, H., Casalicchio, E., & Boeva, V. (2019) · Lecture notes in computer science · doi:10.1007/978-3-030-13453-2_20 · Cited by 31
FATE: fast and accurate timing error prediction framework for low power DNN accelerator design
Zhang, J. J., & Garg, S. (2018) · doi:10.1145/3240765.3240809 · Cited by 27
“Techniques such as zero-skipping and neural network compression seek to further reduce the energy consumption of deep learning accelerators by exploiting special properties of DNNs [1, 5, 17, 21].”
BackgroundManna: An Accelerator for Memory-Augmented Neural Networks
Stevens, J. R., Ranjan, A., Das, D., Kaul, B., & Raghunathan, A. (2019) · IEEE/ACM International Symposium on Microarchitecture (MICRO) · doi:10.1145/3352460.3358304 · Cited by 29
QuTiBench: Benchmarking Neural Networks on Heterogeneous Hardware
Blott, M., Halder, L., Leeser, M. E., & Doyle, L. (2019) · ACM Journal on Emerging Technologies in Computing Systems · doi:10.1145/3358700 · Cited by 33
Deep In-Memory Architectures for Machine Learning–Accuracy Versus Efficiency Trade-Offs
Kang, M., Kim, Y., Patil, A. D., & Shanbhag, N. R. (2020) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2019.2960841 · Cited by 26
F-DNA: Fast Convolution Architecture for Deconvolutional Network Acceleration
Mao, W., Lin, J., & Wang, Z. (2020) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2020.3000519 · Cited by 25
Dadu-Eye: A 5.3 TOPS/W, 30 fps/1080p High Accuracy Stereo Vision Accelerator
Feng, M., Xu, H., Wang, Y., Wang, Y., Li, J., Zou, X., Li, B., & Han, Y. (2021) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2021.3101296 · Cited by 25
Kung, H. T., McDanel, B., & Zhang, S. Q. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1811.04770 · Cited by 25
“These systems, such as the Google TPU [27], the ShiDianNao accelerators [17], and numerous other efforts, including [40], [10], [61], [68], [63], have achieved low power consumption and high throughput.”
BackgroundFLEX-CIM: A Flexible Kernel Size 1-GHz 181.6-TOPS/W 25.63-TOPS/mm 2 Analog Compute-in-Memory Macro
Fu, Y., Yu, W.-H., Un, K.-F., Chan, C., Zhu, Y., Zhang, M., Martins, R. P., & Mak, P. (2024) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2024.3386192 · Cited by 22
“Most power consumed in conventional MAC [4], [5], [6], [7] accelerators are from memory access operations.”
BackgroundYin, X., Huang, Q., Barkam, H. E., Müller, F., Deng, S., Vardar, A., De, S., Jiang, Z., Imani, M., Schlichtmann, U., et al. (2024) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2024.3492994 · Cited by 23
Kılınç, H. Ç., Apak, S., Ergin, M. E., Özkan, F., Katipoğlu, O. M., & Yurtsever, A. (2025) · Acta Geophysica · doi:10.1007/s11600-025-01570-0 · Cited by 21
“CNN may be employed for data feature extraction and data prediction analysis utilizing one-dimensional time series data (Chen et al. 2017).”
Towards Memory Friendly Long-Short Term Memory Networks (LSTMs) on Mobile GPUs
Zhang, X., Xie, C., Wang, J., Zhang, W., & Fu, X. (2018) · IEEE/ACM International Symposium on Microarchitecture (MICRO) · doi:10.1109/micro.2018.00022 · Cited by 26
EFFORT: Enhancing Energy Efficiency and Error Resilience of a Near-Threshold Tensor Processing Unit
Gundi, N. D., Shabanian, T., Basu, P., Pandey, P., Roy, S., Chakraborty, K., & Zhang, Z. (2020) · Asia and South Pacific Design Automation Conference (ASP-DAC) · doi:10.1109/asp-dac47756.2020.9045479 · Cited by 26
“proposed a run-time pruning technique, called row stationary, that enhances the efficiency of a convolutional neural network by re-configuring the spatial architecture, in order to map its computations [24].”
MethodDinelli, G., Meoni, G., Rapuano, E., Pacini, T., & Fanucci, L. (2020) · IEEE Journal on Emerging and Selected Topics in Circuits and Systems · doi:10.1109/jetcas.2020.3015294 · Cited by 22
An In-Flash Binary Neural Network Accelerator with SLC NAND Flash Array
Choi, W. H., Chiu, P.-F., Ma, W., Hemink, G., Hoang, T. T., Lueker-Boden, M., & Bandić, Z. (2020) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas45731.2020.9180920 · Cited by 21
2Deep: Enhancing Side-Channel Attacks on Lattice-Based Key-Exchange via 2-D Deep Learning
Kashyap, P., Aydın, F., Potluri, S., Franzon, P. D., & Aysu, A. (2020) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2020.3038701 · Cited by 21
DeepNVM++: Cross-Layer Modeling and Optimization Framework of Nonvolatile Memories for Deep Learning
Inci, A. F., Isgenc, M. M., & Marculescu, D. (2021) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2021.3127148 · Cited by 28
“We choose batch size 4 for inference and 64 for training for our workloads as it is typically used in related work [63].”
BackgroundMethodLee, J.-J., & Li, P. (2020) · IEEE International Conference on Computer Design: VLSI in Computers and Processors (ICCD) · doi:10.1109/iccd50377.2020.00027 · Cited by 28
“We focus on the accelerating CONV layers as they account for over 90% of total computation [3, 21-22].”
BackgroundMethodInfluentialWu, Y. N., Tsai, P.-A., Parashar, A., Sze, V., & Emer, J. (2021) · IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS) · doi:10.1109/ispass51385.2021.00043 · Cited by 21
Self-citationHigh performance accelerators for deep neural networks: A review
Akhoon, M. S., Suandi, S. A., Alshahrani, A. S., Saad, A. H. Y., Albogamy, F. R., Abdullah, M. Z., & Loan, S. A. (2021) · Expert Systems · doi:10.1111/exsy.12831 · Cited by 23
“Chen et al. (2017, 2019) have designed and realized Eyeriss and Eyeriss V2 DNN accelerator architectures.”
BackgroundMethodEmerging Energy-Efficient Biosignal-Dedicated Circuit Techniques: A Tutorial Brief
Zhao, S., Fang, C., Yang, J., & Sawan, M. (2022) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2022.3169004 · Cited by 25
“carry out the computation, data moves back and forth between the memory and the arithmetic unit, causing a lot of energy loss [68].”
BackgroundTortorella, Y., Bertaccini, L., Rossi, D., Benini, L., & Conti, F. (2022) · Design, Automation, and Test in Europe · doi:10.23919/date54114.2022.9774759 · Cited by 24
uBrain: a unary brain computer interface
Wu, D., Li, J., Pan, Z., Kim, Y., & Miguel, J. S. (2022) · Proceedings - International Symposium on Computer Architecture/Proceedings · doi:10.1145/3470496.3527401 · Cited by 28
APTPU: Approximate Computing Based Tensor Processing Unit
Elbtity, M., Chandarana, P. S., Reidy, B., Eshraghian, J. K., & Zand, R. (2022) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2022.3206262 · Cited by 23
TelaMalloc: Efficient On-Chip Memory Allocation for Production Machine Learning Accelerators
Maas, M., Beaugnon, U., Chauhan, A., & Ilbeyi, B. (2022) · International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS) · doi:10.1145/3567955.3567961 · Cited by 42
“Examples of such accelerators include the widely studied EyeRiss accelerator [24] as well as the Timeloop [51] and MAESTRO [35] models.”
MethodNon-uniform Step Size Quantization for Accurate Post-training Quantization
Oh, S., Sim, H., Kim, J., & Lee, J. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-20083-0_39 · Cited by 33
“Then a decoder is practically reduced to a simple MUX, but even this MUX can often be shared among a number of MACs within a MAC array, depending on the dataflow of the MAC array [5].”
BackgroundAn Energy-Efficient Accelerator for Medical Image Reconstruction From Implicit Neural Representation
Rao, C., Wu, Q., Zhou, P., Jingyi Yu, A., Zhang, Y., & Lou, X. (2022) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2022.3231863 · Cited by 22
Kim, S., Jang, B., Lee, J., Bae, H., Jang, H., & Park, I. (2023) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2023.3234640 · Cited by 25
Tan, J., Wang, Q., Yan, K., Wei, X., & Fu, X. (2023) · Future Generation Computer Systems · doi:10.1016/j.future.2023.05.009 · Cited by 29
Feng, K., Kong, T., Koul, K., Melchert, J., Carsello, A., Liu, Q., Nyengele, G., Strange, M., Zhang, K., Nayak, A., et al. (2023) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2023.3313116 · Cited by 29
Security of Electrical, Optical, and Wireless On-chip Interconnects: A Survey
Weerasena, H., & Mishra, P. (2023) · ACM Transactions on Design Automation of Electronic Systems · doi:10.1145/3631117 · Cited by 21
“Novel dataflow-based DNN accelerators use application-specific on-chip communication fabric to transfer weights, activations and partial sums among processing cores and buffers [32, 33, 58].”
On-Chip Networks, Second Edition
Jerger, N. E., Krishna, T., & Peh, L.-S. (2017) · Synthesis lectures on computer architecture · doi:10.2200/s00772ed1v01y201704cac040 · Cited by 58
Self-citationTensorQuant: A Simulation Toolbox for Deep Neural Network Quantization
Loroch, D. M., Pfreundt, F., Wehn, N., & Keuper, J. (2017) · doi:10.1145/3146347.3146348 · Cited by 27
“max_pool2d(net , [2, 2], 2, scope='pool1 ')”
MethodModified distributed arithmetic based low complexity CNN architecture design methodology
Panwar, M., Padmini, J., Venkatasubrahmanian, Acharyya, A., & Biswas, D. (2017) · European Conference on Circuit Theory and Design · doi:10.1109/ecctd.2017.8093254 · Cited by 23
“Several researchers have focused on designing a less complex, low area and low-power designs to enable deployment of CNNs in embedded devices [16-17].”
BackgroundFeedforward-Cutset-Free Pipelined Multiply–Accumulate Unit for the Machine Learning Accelerator
Ryu, S., Park, N., & Kim, J. (2018) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2018.2873716 · Cited by 22
“Since an enormous amount of vector-matrix multiplication computations are required in a typical DNN application, a variety of dedicated hardware for machine learning have been proposed to accelerate the computations [6], [7].”
BackgroundMethodInstruction Driven Cross-layer CNN Accelerator for Fast Detection on FPGA
Yu, J., Ge, G., Hu, Y., Ning, X., Qiu, J., Guo, K., Wang, Y., & Yang, H. (2018) · ACM Transactions on Reconfigurable Technology and Systems · doi:10.1145/3283452 · Cited by 24
“Previous works [5, 20] that utilize off-chip memory discuss the scheduling or data arrangement strategy to overcome the bandwidth limitation between the external and the on-chip memory.”
BackgroundMethodOptimize Deep Convolutional Neural Network with Ternarized Weights and High Accuracy
He, Z., Gong, B., & Fan, D. (2019) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv.2019.00102 · Cited by 27
GenUnlock: An Automated Genetic Algorithm Framework for Unlocking Logic Encryption
Chen, H., Fu, C., Zhao, J., & Koushanfar, F. (2019) · doi:10.1109/iccad45719.2019.8942134 · Cited by 23
“Integrated circuits (ICs) are indispensable for various real-world applications ranging from domestic electronics, autonomous vehicles to medical devices and deep learning systems [6, 17].”
BackgroundThe Development of Silicon for AI: Different Design Approaches
Lee, K. J., Lee, J., Choi, S., & Yoo, H. (2020) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2020.2996625 · Cited by 24
“Chen et al [35] proposed a CNN accelerator utilizing data-reuse pattern and its dedicated hardware architecture, and Sim et al [36] also designed a digital CNN processor with multi-range MAC unit”
MethodVisibility Restoration: A Systematic Review and Meta-Analysis
Ngo, D., Lee, S., Ngo, T. M., Lee, G., & Kang, B. (2021) · Sensors · doi:10.3390/s21082625 · Cited by 23
“In the literature, Eyeriss [109] and its successor [110] provided an energy-efficient framework for designing deep neural networks.”
BackgroundWang, X., Feng, C., Kang, X., Wang, Q., Huang, Y., & Ye, T. T. (2024) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2024.3472293 · Cited by 21
A Survey of Artificial Neural Network Computing Systems
Foukalas, F. T. (2024) · Cognitive Computation · doi:10.1007/s12559-024-10383-0 · Cited by 30
Lou, Q., Gao, T., Faley, P., Niemier, M. T., Hu, X. S., & Joshi, S. (2020) · International Symposium on Low Power Electronics and Design (ISLPED) · doi:10.1145/3370748.3406583 · Cited by 23
“For software trained DNNs to be reliably deployed onto a large number of diverse hardware platforms, the hardware must be able to deliver accurate results despite inherent unreliability in the underlying components.”
BackgroundInfluentialEfficient Accelerator for Dilated and Transposed Convolution with Decomposition
Chang, K.-W., & Chang, T. (2020) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas45731.2020.9180402 · Cited by 20
“Various hardware accelerators [10]–[15] have been proposed recently.”
BackgroundAn Architecture-Level Energy and Area Estimator for Processing-In-Memory Accelerator Designs
Wu, Y. N., Sze, V., & Emer, J. (2020) · IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS) · doi:10.1109/ispass48437.2020.00024 · Cited by 25
“In recent years, many DNN accelerators have been proposed [1]–[8] to improve the energy efficiency of DNN applications by exploiting the application-specific properties.”
BackgroundSelf-citationA Hybrid Computing Architecture for Fault-tolerant Deep Learning Accelerators
Xu, D., Chu, C., Wang, Q., Liu, C., Wang, Y., Zhang, L., Liang, H., & Cheng, K.-T. T. (2020) · IEEE International Conference on Computer Design: VLSI in Computers and Processors (ICCD) · doi:10.1109/iccd50377.2020.00087 · Cited by 21
“DLAs typically consist of a large regular computing array which can either be a systolic array or a plain mesh array [7] [8], and a set of on-chip buffers used for input/output feature data and weights.”
Learning Without Forgetting: A New Framework for Network Cyber Security Threat Detection
Karn, R. R., Kudva, P. N., & Elfadel, I. M. (2021) · IEEE Access · doi:10.1109/access.2021.3115946 · Cited by 32
A Deep Learning Accelerator Based on a Streaming Architecture for Binary Neural Networks
Vo, Q. H., Le, N. L., Asim, F., Kim, L.-W., & Hong, C. S. (2022) · IEEE Access · doi:10.1109/access.2022.3151916 · Cited by 23
GNNIE: GNN inference engine with load-balancing and graph-specific caching
Mondal, S., Manasi, S. D., Kunal, K., Ramprasath, S., & Sapatnekar, S. S. (2022) · ACM/IEEE Design Automation Conference (DAC) · doi:10.1145/3489517.3530503 · Cited by 29
“convolutional neural network (CNN) computations, but CNN accelerators [2], [7], [15], [17], [18], [27], [30] are inefficient at handling graph data.”
BackgroundJu, Y., & Gu, J. (2022) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2022.3214170 · Cited by 24
“explored at the software level such as quantization [1], architecture level such as flexible dataflows [2], and bit-precision”
BackgroundA 4-Bit Mixed-Signal MAC Macro With One-Shot ADC Conversion
Yang, X., & Sun, N. (2023) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2023.3269076 · Cited by 23
“While digital accelerators have shown high energy efficiency [5], [6], [7], the mixed-signal charge-domain MAC circuits combined with in-memory-computing (IMC) have demonstrated the potential to outperform their digital counterparts [8], [9].”
Dorrance, R., Dasalukunte, D., Wang, H., Liu, R., & Carlton, B. R. (2023) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2023.3283186 · Cited by 25
“Dominated by highly parallel matrix-vector multiplications (MVMs), NN-based machine learning inference accelerators have made several attempts to address the so-called “memory wall” introduced by the frequent data movement between the computing elements and the memory units [8], [9], [10], [11].”
BackgroundThe Benefits and Risks of Artificial General Intelligence (AGI)
Fahad, M., Basri, T., Hamza, M. A., Faisal, S., Akbar, A., Haider, U., & Hajjami, S. E. (2024) · Advanced technologies and societal change · doi:10.1007/978-981-97-3222-7_2 · Cited by 33
Luo, X., Liu, D., Kong, H., Huai, S., Chen, H., Xiong, G., & Liu, W. (2024) · ACM Transactions on Embedded Computing Systems · doi:10.1145/3701728 · Cited by 33
Chen, X., Chen, Q., Tian, M., Ji, M., Zou, C., Wang, X., & Wang, B. (2017) · IEEE International Conference on Parallel and Distributed Systems (ICPADS) · doi:10.1109/icpads.2017.00034 · Cited by 23
“Eyeriss [20] is one of the most energy-efficient state-of-the-art hardware architecture.”
BackgroundChen, C., Ding, H., Peng, H., Zhu, H., Wang, Y., & Shi, C. ‐. R. (2018) · IEEE Journal on Emerging and Selected Topics in Circuits and Systems · doi:10.1109/jetcas.2018.2852780 · Cited by 21
Lee, D., Kang, S.-B., & Choi, K. (2018) · International Conference on Supercomputing · doi:10.1145/3205289.3205295 · Cited by 27
“Typical DNNs require a lot of computing power and energy, and to alleviate the problem, various accelerators such as [1, 2, 4–6, 9, 12] have been developed to execute DNNs more efficiently.”
MethodPAGURUS: Low-Overhead Dynamic Information Flow Tracking on Loosely Coupled Accelerators
Piccolboni, L., Guglielmo, G. D., & Carloni, L. P. (2018) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2018.2857321 · Cited by 22
“been developed for different applications, including neural networks [5], [6], database processing [7], [8], graph pro-”
Li, Y., Liu, Z., Liu, W., Jiang, Y. S., Wang, Y., Goh, W. L., Yu, H., & Ren, F. (2018) · IEEE Transactions on Industrial Electronics · doi:10.1109/tie.2018.2875643 · Cited by 24
Zhang, B., Han, J. S., Huang, Z., Yang, J., & Zeng, X. (2019) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2019.2899829 · Cited by 23
“The Eyeriss processor [33] performs VGG-16 at 0.”
MethodCause identification of electromagnetic transient events using spatiotemporal feature learning
Niazazari, I., Hamidi, R. J., Livani, H., & Arghandeh, R. (2020) · International Journal of Electrical Power & Energy Systems · doi:10.1016/j.ijepes.2020.106255 · Cited by 25
Zhang, J. J., Raj, P., Zarar, S., Ambardekar, A., & Garg, S. (2019) · ACM Transactions on Embedded Computing Systems · doi:10.1145/3358178 · Cited by 21
“To mitigate these issues, there is growing interest in the design of hardware accelerators to speed-up CNN inference from both industry and academia [4, 5, 9, 17, 27, 30].”
BackgroundMethodInfluentialWang, Y., & Li, H. (2020) · Computers & Chemical Engineering · doi:10.1016/j.compchemeng.2020.106877 · Cited by 21
Accelerating Deep Convolutional Neural Network base on stochastic computing
Sadi, M. H., & Mahani, A. K. (2020) · Integration · doi:10.1016/j.vlsi.2020.09.008 · Cited by 21
Xu, C., Luong, T. V., Xiang, L., Sugiura, S., Maunder, R. G., Yang, L., & Hanzo, L. (2022) · IEEE Transactions on Cognitive Communications and Networking · doi:10.1109/tccn.2022.3168725 · Cited by 28
An Asynchronous Energy-Efficient CNN Accelerator with Reconfigurable Architecture
Chen, W., Wu, H., Wei, S., He, A., & Chen, H. (2018) · IEEE Asian Solid-State Circuits Conference (A-SSCC) · doi:10.1109/asscc.2018.8579304 · Cited by 21
“To obtain a better flexibility and energy efficiency, some CNN accelerators adopt a coarse-grained dynamic reconfigurable architecture (CGRA) such as the Eyeriss from MIT [4] and the Thinker [ 1 ] with high performance and flexibility.”
MethodTaurus: An Intelligent Data Plane.Preprint
Swamy, T., Rucker, A., Shahbaz, M., & Olukotun, K. (2020) · arXiv (Cornell University) · Cited by 21
Boroumand, A., Ghose, S., Akin, B., Narayanaswami, R., Junior, G. F. D. O., Ma, X., Shiu, E., & Mutlu, O. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2103.00768 · Cited by 41
“Many prior works [1,2,5,6,7,13,15,22,23,33,39,42,45,46, 48, 50, 51, 57] look at a specific model type (predominantly CNNs).”
BackgroundFirst-Generation Inference Accelerator Deployment at FacebookPreprint
Anderson, M. J., Chen, B., Chen, S. Y., Deng, S., Fix, J., Gschwind, M. K., Kalaiah, A., Kim, C., Lee, J., Liang, J., et al. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2107.04140 · Cited by 35
“CNN/DNNs [8], [10], [21], MANNs [27], [47], [58], recommendation models [25], an arbitrary DNN using a reconfigurable hardware or a systematic methodology [35], [36], [51], [56], and generic ISAs [38] have been proposed.”
MethodFLightNNs: Lightweight Quantized Deep Neural Networks for Fast and Accurate Inference
Ding, R., Liu, Z., Chin, T.-W., Marculescu, D., & Blanton, R. D. (. (2019) · Proceedings of the 56th Annual Design Automation Conference 2019 · doi:10.1145/3316781.3317828 · Cited by 19
Retrain-Less Weight Quantization for Multiplier-Less Convolutional Neural Networks
Choi, J., Kong, B. Y., & Park, I. (2019) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2019.2949935 · Cited by 19
A Throughput-Optimized Channel-Oriented Processing Element Array for Convolutional Neural Networks
Yu-xian, C., & Ruan, S.-J. (2020) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2020.3017789 · Cited by 20
“This strategy is included in a mapping rule called a dataflow [8].”
MethodMiniaturization for wearable EEG systems: recording hardware and data processing
Kim, M., Yoo, S.-J., & Kim, C. (2022) · Biomedical Engineering Letters · doi:10.1007/s13534-022-00232-0 · Cited by 25
“To optimize memory-related data flow, the tiling of for-looped operations is focused [113, 114].”
MethodLou, J., Freye, F., Lanius, C., & Gemmeke, T. (2023) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2023.3323205 · Cited by 22
“Data movement consumes more energy than arithmetic operations [8], [9].”
BackgroundFPGA-Based CNN for Eye Detection in an Iris Recognition at a Distance System
Ruiz-Beltrán, C. A., Romero‐Garcés, A., González-García, M., Marfil, R., & Bandera, A. J. (2023) · Electronics · doi:10.3390/electronics12224713 · Cited by 24
Compute-In-Memory Technologies for Deep Learning Acceleration
Meng, F.-H., & Lu, W. D. (2024) · IEEE Nanotechnology Magazine · doi:10.1109/mnano.2023.3340321 · Cited by 23
“Hardwares designed specifically to accelerate deep neural networks (DNNs) have become a hot topic of interest in the past decade due to the success of deep learning [1], [2], [13], [14], [15].”
BackgroundFlexNPU: a dataflow-aware flexible deep learning accelerator for energy-efficient edge devices
Raha, A., Mathaikutty, D. A., Kundu, S., & Ghosh, S. K. (2025) · Frontiers in High Performance Computing · doi:10.3389/fhpcp.2025.1570210 · Cited by 37
“…in adapting to the rapidly evolving landscape of DNNs. Existing DNN accelerator designs from both industry and academia predominantly employ fixed schedules, such as input stationary (IS), weight stationary (WS), output stationary (OS), non-local reuse (NR), and row stationary (RS) [8], [9].”
Armeniakos, G., Maras, A., Xydis, S., & Soudris, D. (2024) · doi:10.1145/3676536.3676840 · Cited by 30
“These accelerators often provide significant performance enhancements, but those benefits are often compromised due to their lack of scalability, extensive area occupation, and high power consumption, typically in the range of hundreds of milliwatts [6, 13, 24] to a few watts [25].”
High-Throughput and Energy-Efficient FPGA-Based Accelerator for All Adder Neural Networks
Zhang, N., Ni, S., Chen, L., Wang, T., & Chen, H. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3543213 · Cited by 26
CoNNA – Compressed CNN Hardware Accelerator
STRUHARIK, R. J. R., Vukobratović, B. Z., Erdeljan, A., & Rakanovic, D. (2018) · Euromicro Conference on Digital System Design (DSD) · doi:10.1109/dsd.2018.00070 · Cited by 20
“For example, this approach is used in MIT’s Eyeriss accelerator [12].”
BackgroundMethodInfluentialScalable Distributed Last-Level TLBs Using Low-Latency Interconnects
Bharadwaj, S., Cox, G., Krishna, T., & Bhattacharjee, A. (2018) · IEEE/ACM International Symposium on Microarchitecture (MICRO) · doi:10.1109/micro.2018.00030 · Cited by 21
Self-citationKim, C., Shin, D., Kim, B., & Park, J. (2018) · IEEE Journal on Emerging and Selected Topics in Circuits and Systems · doi:10.1109/jetcas.2018.2865006 · Cited by 19
Customized High Performance and Energy Efficient Communication Networks for AI Chips
Gao, W., & Zhou, P. (2019) · IEEE Access · doi:10.1109/access.2019.2916338 · Cited by 19
Early-Stage Neural Network Hardware Performance Analysis
Karbachevsky, A., Baskin, C., Zheltonozhskii, E., Yermolin, Y., Gabbay, F., Bronstein, A. M., & Mendelson, A. (2021) · Sustainability · doi:10.3390/su13020717 · Cited by 21
“While two of these layers are close to the memory bound and can possibly get inside the memory bound of the compression scheme [20], the first and the three last layers suffer from poor performance compared to other layers.”
BackgroundMethodInfluentialManasi, S. D., Snigdha, F. S., & Sapatnekar, S. S. (2020) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2020.2995135 · Cited by 22
Riaz, M., Hafiz, R., Khaliq, S. A., Faisal, M., Iqbal, H. T., Ali, M., & Shafique, M. (2020) · IEEE Access · doi:10.1109/access.2020.3008256 · Cited by 24
Sateesan, A., Sinha, S., Smitha, K. G., & Vinod, A. P. (2021) · Neural Processing Letters · doi:10.1007/s11063-021-10458-1 · Cited by 22
“AnacceleratorarchitectureEyeriss[89]wasproposedtodeliverstateoftheartperfor-mancewithminimumenergyconsumption.”
BackgroundInfluentialExtending Sparse Tensor Accelerators to Support Multiple Compression Formats
Qin, E., Jeong, G., Won, W., Kao, S.-C., Kwon, H., Srinivasan, S., Das, D., Moon, G. E., Rajamanickam, S., & Krishna, T. (2021) · IEEE International Parallel and Distributed Processing Symposium (IPDPS) · doi:10.1109/ipdps49936.2021.00110 · Cited by 24
Self-citationLi, R., Huang, P. C., Feng, Y., Zhou, Z., Zhang, Y., Ding, X., Liu, L., & Kang, J. (2022) · Micromachines · doi:10.3390/mi13030433 · Cited by 20
A 200MHZ 202.4GFLOPS@10.8W VGG16 accelerator in Xilinx VX690T
Mei, C., Liu, Z., Niu, Y., Ji, X., Zhou, W., & Wang, D. (2017) · IEEE Global Conference on Signal and Information Processing · doi:10.1109/globalsip.2017.8309067 · Cited by 28
“Because the uneven computational overhead among different convolutional layers and and the data dependency between adjacent layers, the up-to-date CNN accelerator designs usually apply reconfigurable processing units with tiled input feature map to carry out all convolutions [3–5, 7, 13].”
BackgroundAddressing Sparsity in Deep Neural Networks
Zhou, X., Chen, Y., Du, Z., Zhang, S., Zhang, L., Lan, H., Liu, S., Li, L., Guo, Q., & Chen, T. (2018) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2018.2864289 · Cited by 21
“Eyeriss [48] applies run-length compression and PE data gating to exploit zeros in the feature maps (activation sparsity), thus saving DRAM bandwidth and PE energy consumption.”
MethodSelf-citation
Metadata
- Authors
- Chen, Y.-H., Krishna, T., Emer, J. S., & Sze, V.
- Affiliations
- Massachusetts Institute of Technology (US)
- Georgia Institute of Technology (US)
- Nvidia (United States) (US)
- Venue
- IEEE Journal of Solid-State Circuits
- Published
- 2017-01
- Volume
- 52
- Issue
- 1
- Pages
- 127-138
- ISSN
- 0018-9200 (ISSN-L)
- 1558-173X
- OpenAlex
- W2289252105
- Semantic Scholar
- ffdaa12ef011de9dbf43be46d45a3abcc8288965
- Cited by
- 3,247
- References
- 42
- Citations by year
- 2026: 218 · 2025: 313 · 2024: 294 · 2023: 373 · 2022: 365 · 2021: 450 · 2020: 427 · 2019: 354 · 2018: 302 · 2017: 118
- FWCI
- 182.12
- Type
- journal article
- Language
- en
- Open access
- closed
- Locations
- IEEE Journal of Solid-State Circuits (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Hardware and Architecture › Parallel Computing and Optimization Techniques
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Advanced Memory and Neural Computing
- Keywords
- reconfigurable accelerators, deep convolutional neural networks, energy efficiency, spatial architecture, data movement reduction, data compression, processing elements, AlexNet, VGG16, convolutional layers
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.