CONV-SRAM: An Energy-Efficient SRAM With In-Memory Dot-Product Computation for Low-Power Convolutional Neural Networks
Open accessThe DOI matches its Crossref record; 25 of the 29 references it cites were found in a registry or an index.
Abstract
This paper presents an energy-efficient static random access memory (SRAM) with embedded dot-product computation capability, for binary-weight convolutional neural networks. A 10T bit-cell-based SRAM array is used to store the 1-b filter weights. The array implements dot-product as a weighted average of the bitline voltages, which are proportional to the digital input values. Local integrating analog-to-digital converters compute the digital convolution outputs, corresponding to each filter. We have successfully demonstrated functionality (>98% accuracy) with the 10 000 test images in the MNIST hand-written digit recognition data set, using 6-b inputs/outputs. Compared to conventional full-digital implementations using small bitwidths, we achieve similar or better energy efficiency, by reducing data transfer, due to the highly parallel in-memory analog computations.
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
“Among charge-based memories, SRAM cells (a bi-stable transistor structure typically made of two CMOS inverters connected back-to-back which retains a charge concentration, see Fig. 3a for an example of the structure of a crossbar array of 6T SRAM) have been widely used for VMM75–77.”
From Hardware implementation of memristor-based artificial neural networks (2024) · cites this paper as [76]
Citations
Cites
ImageNet 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 · cited 1×
“Convolutional neural networks (CNNs) provide state-of-theart results in a wide variety of AI/ML applications, ranging from image classification [3] to speech recognition [1].”
BackgroundGradient-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 5×
“However, for the MNIST dataset with LeNet-5 CNN, we estimate [22] those to have only small contributions to the overall energy-efficiency per MAV operation, due to the high parallelism supported by our in-memory approach.”
MethodInfluentialHinton, G., Deng, L., Yu, D., Dahl, G., Mohamed, A.-R., Jaitly, N., Senior, A., Vanhoucke, V., Nguyen, P., Sainath, T., et al. (2012) · IEEE Signal Processing Magazine · doi:10.1109/msp.2012.2205597 · Cited by 10,260
DeepFace: Closing the Gap to Human-Level Performance in Face Verification
Taigman, Y., Yang, M., Ranzato, M., & Wolf, L. (2014) · 2014 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2014.220 · Cited by 6,542 · cited 1×
“Facial recognition [2] enables using our faces to get access to devices in a more intuitive manner, instead of traditional passcodes.”
BackgroundXNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks
Rastegari, M., Ordonez, V., Redmon, J., & Farhadi, A. (2016) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-46493-0_32 · Cited by 4,408 · cited 2×
“[6] proposed a binary-weight-network (BWN), where the filter weights (wi s) can be trained to be +1/−1 (with a common scaling factor per 3-D filter: α).”
BackgroundMethodEfficient Processing of Deep Neural Networks: A Tutorial and Survey
Sze, V., Chen, Y.-H., Yang, T.-J., & Emer, J. S. (2017) · Proceedings of the IEEE · doi:10.1109/jproc.2017.2761740 · Cited by 4,215 · cited 1×
“In the conventional all-digital implementation of CNNs [4], [5], [9], with the memory and the processing elements being physically separate, reading wi s and the partial sums from the on-chip SRAMs lead to a lot of data movement per computation [10] and, hence, make them energy-hungry.”
BackgroundEyeriss: An Energy-Efficient Reconfigurable Accelerator for Deep Convolutional Neural Networks
Chen, Y.-H., Krishna, T., Emer, J. S., & Sze, V. (2017) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2016.2616357 · Cited by 3,247 · cited 12×
“CNNs typically consist of a cascade of convolutional (CONV) and fully-connected (FC) layers (Fig.”
BackgroundMethodInfluentialBinaryConnect: Training Deep Neural Networks with binary weights during propagations
Courbariaux, M., Bengio, Y., & David, J.-P. (2015) · arXiv · doi:10.48550/arxiv.1511.00363 · Cited by 2,180 · cited 1×
“BWN performs better than binary connect [7], which does not incorporate the scaling factor of α per filter, and also binarized neural networks [8], where both weights and activations are constrained to ±1.”
Background1.1 Computing's energy problem (and what we can do about it)
Horowitz, M. (2014) · 2014 IEEE International Solid-State Circuits Conference Digest of Technical Papers (ISSCC) · doi:10.1109/isscc.2014.6757323 · Cited by 2,069 · cited 1×
“to access data from memory can be much higher than the energy needed for a compute operation with that data [11].”
BackgroundHubara, I., Soudry, D., & Yaniv, R. E. (2016) · arXiv · doi:10.48550/arxiv.1602.02505 · Cited by 1,667 · cited 1×
“BWN performs better than binary connect [7], which does not incorporate the scaling factor of α per filter, and also binarized neural networks [8], where both weights and activations are constrained to ±1.”
Background
Research works printed but not linked: 1
- whatmough, 2017, A 28 nm SoC with a 1.2 GHz 568 nJ/prediction sparse deep-neural-network engine with >0.1 timing error rate tolerance for IoT applications, IEEE Int Solid-State Circuits Conf (ISSCC) Dig Tech Papers, 242
Cited by
Memory devices and applications for in-memory computing
Sebastian, A., Le Gallo, M., Khaddam-Aljameh, R., & Eleftheriou, E. (2020) · Nature Nanotechnology · doi:10.1038/s41565-020-0655-z · Cited by 2,289
Edge Computing with Artificial Intelligence: A Machine Learning Perspective
Hua, H., Li, Y., Wang, T., Dong, N., Li, W., & Cao, J. (2023) · ACM Computing Surveys · doi:10.1145/3555802 · Cited by 563
“A static random access memory (SRAM) [102] is able to reduce memory data throughput, and it combines parallel CNNs to enable simultaneous access to different memory blocks.”
BackgroundMethodInfluentialHardware implementation of memristor-based artificial neural networks
Aguirre, F., Sebastian, A., Le Gallo, M., Song, W., Wang, T., Yang, J. J., Lu, W., Chang, M.-F., Ielmini, D., Yang, Y., et al. (2024) · Nature Communications · doi:10.1038/s41467-024-45670-9 · Cited by 564
“The proposed circuit comprises 4 kb synapses fabricated in a 65 nm CMOS processand reported anenergy ef fi ciencyof 55.8 TOPS per W. Incases where x is non-binary, one approach is to employ capacitors in addition to the SRAM cells 76,77,79 , involving a three-step process.”
InfluentialIn-Memory Computing: Advances and Prospects
Verma, N., Jia, H., Valavi, H., Tang, Y., Ozatay, M., Chen, L.-Y., Zhang, B., & Deaville, P. (2019) · IEEE Solid-State Circuits Magazine · doi:10.1109/mssc.2019.2922889 · Cited by 473
“, [13] requires adopting a 10-T bit cell, and [14] requires multiple memory-operation cycles).”
BackgroundXNOR-SRAM: In-Memory Computing SRAM Macro for Binary/Ternary Deep Neural Networks
Yin, S., Jiang, Z., Seo, J.-S., & Seok, M. (2020) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2019.2963616 · Cited by 429
“For example, CONV-SRAM [20] integrates digital-toanalog converters (DACs) for analog wordlines, binary weights stored in SRAM, and analog-to-digital converters (ADCs) to convert the in-memory computation results back to digital values.”
BackgroundMethodC3SRAM: An In-Memory-Computing SRAM Macro Based on Robust Capacitive Coupling Computing Mechanism
Jiang, Z., Yin, S., Seo, J.-S., & Seok, M. (2020) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2020.2992886 · Cited by 324
“such as [21] and Conv-SRAM [22].”
BackgroundMethodResultInfluentialLe Gallo, M., Khaddam-Aljameh, R., Stanisavljevic, M., Vasilopoulos, A., Kersting, B., Dazzi, M., Karunaratne, G., Brändli, M., Singh, A., Müller, S. M., et al. (2023) · Nature Electronics · doi:10.1038/s41928-023-01010-1 · Cited by 310
“Although the SRAM-based accelerator presented in another work 24 has a higher energy efficiency as well as throughput density, it requires the use of an off-chip weight buffer to store the total weight data of a network, leading to additional The LSTM unit comprises an input gate that receives the embedded input chara…”
BackgroundInfluentialWang, J., Wang, X., Eckert, C., Subramaniyan, A., Das, R., Blaauw, D., & Sylvester, D. (2020) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2019.2939682 · Cited by 244
“In this case, one operand is pre-stored in the SRAM array, while the other operand can be modulated into the word-line voltage level [14] or pulsewidth of the word-line enable signal [15], [16].”
BackgroundMethodA Twin-8T SRAM Computation-in-Memory Unit-Macro for Multibit CNN-Based AI Edge Processors
Si, X., Chen, J.-J., Tu, Y.-N., Huang, W.-H., Wang, J.-H., Chiu, Y.-C., Wei, W.-C., Wu, S.-Y., Sun, X., Liu, R., et al. (2020) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2019.2952773 · Cited by 231
Kim, H., Yoo, T., Kim, T. T.-H., & Kim, B. (2021) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2021.3061508 · Cited by 199
Si, X., Tu, Y.-N., Huang, W.-H., Su, J.-W., Lu, P.-J., Wang, J., Liu, T.-W., Wu, S.-Y., Liu, R., Chou, Y.-C., et al. (2021) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2021.3073254 · Cited by 161
A full spectrum of computing-in-memory technologies
Sun, Z., Kvatinsky, S., Si, X., Mehonić, A., Cai, Y., & Huang, R. (2023) · Nature Electronics · doi:10.1038/s41928-023-01053-4 · Cited by 199
Xie, S., Ni, C., Sayal, A., Jain, P., Hamzaoglu, F., & Kulkarni, J. P. (2021) · Digest of technical papers/Digest of technical papers - IEEE International Solid-State Circuits Conference · doi:10.1109/isscc42613.2021.9365932 · Cited by 154
Khaddam-Aljameh, R., Stanisavljevic, M., Fornt Mas, J., Karunaratne, G., Brändli, M., Liu, F., Singh, A., Müller, S. M., Egger, U., Petropoulos, A., et al. (2022) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2022.3140414 · Cited by 144
Chen, Z., Yu, Z., Jin, Q., He, Y., Wang, J., Lin, S., Li, D., Wang, Y., & Yang, K. (2021) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2021.3056447 · Cited by 148
“A serial ADC [11], which uses a single dummy cell as a reference, achieves 7-bit resolution and high area-efficiency, but the serial nature results in low throughput.”
BackgroundMethodInfluentialWang, Y., Tang, H., Xie, Y., Chen, X., Ma, S., Sun, Z., Sun, Q., Chen, L., Zhu, H., Wan, J., et al. (2021) · Nature Communications · doi:10.1038/s41467-021-23719-3 · Cited by 152
Wu, P., He, T., Zhu, H. L., Wang, Y., Li, Q., Wang, Z., Fu, X. D., Wang, F., Wang, P., Shan, C., et al. (2021) · InfoMat · doi:10.1002/inf2.12275 · Cited by 134
“31 A near-sensor architecture adopting near-memory computing has been proposed that shortens the transmission distance between storage and computation and improves energy efficiency in traditional-film-material systems.”
BackgroundHung, J.-M., Xue, C.-X., Kao, H.-Y., Huang, Y.-H., Chang, F.-C., Huang, S.-P., Liu, T.-W., Jhang, C.-J., Su, C.-I., Khwa, W.-S., et al. (2021) · Nature Electronics · doi:10.1038/s41928-021-00676-9 · Cited by 133
A 65-nm 8T SRAM Compute-in-Memory Macro With Column ADCs for Processing Neural Networks
Yu, C., Yoo, T., Chai, K. T. C., Kim, T. T.-H., & Kim, B. (2022) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2022.3162602 · Cited by 137
“A 10T SRAM-based bitcell [6] is used to construct a macro”
BackgroundMethodLin, Z., Zhu, Z., Zhan, H., Peng, C., Wu, X., Yao, Y., Niu, J., & Chen, J. (2021) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2021.3061260 · Cited by 97
Roadmap for unconventional computing with nanotechnology
Finocchio, G., Incorvia, J. A. C., Friedman, J. S., Qu, Y., Giordano, A., Grollier, J., Yang, H., Ciubotaru, F., Chumak, A. V., Naeemi, A., et al. (2024) · Nano Futures · doi:10.1088/2399-1984/ad299a · Cited by 137
An Overview of Processing-in-Memory Circuits for Artificial Intelligence and Machine Learning
Kim, D., Yu, C., Xie, S., Chen, Y., Kim, J.-Y., Kim, B., Kulkarni, J. P., & Kim, T. T.-H. (2022) · IEEE Journal on Emerging and Selected Topics in Circuits and Systems · doi:10.1109/jetcas.2022.3160455 · Cited by 104
“For example, [3] added an 8:1 MUX, 3 2:1 MUX, and extra transistors, [13] realized the flash ADC by adding 15 dedicated sense amplifiers and [12] implemented more than 32 extra transistors to achieve the current DAC design.”
BackgroundMethodInfluentialTD-SRAM: Time-Domain-Based In-Memory Computing Macro for Binary Neural Networks
Song, J., Wang, Y., Guo, M., Ji, X. L., Cheng, K., Hu, Y., Tang, X., Wang, R., & Huang, R. (2021) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2021.3083275 · Cited by 89
PIMCA: A Programmable In-Memory Computing Accelerator for Energy-Efficient DNN Inference
Zhang, B., Yin, S., Kim, M., Saikia, J., Kwon, S., Myung, S., Kim, H., Kim, S. J., Seo, J.-S., & Seok, M. (2022) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2022.3211290 · Cited by 80
“SRAM macros with respect to the computing mechanisms they adopt, namely: 1) current discharging [9], [10], [13], [14], [15],”
BackgroundLee, E., Han, T., Seo, D., Shin, G., Kim, J., Kim, S., Jeong, S., Rhe, J., Park, J., Ko, J. H., et al. (2021) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2021.3080042 · Cited by 80
Time-Domain Computing in Memory Using Spintronics for Energy-Efficient Convolutional Neural Network
Zhang, Y., Wang, J., Lian, C., Bai, Y., Wang, G., Zhang, Z., Zheng, Z., Chen, L., Zhang, K., Sirakoulis, G. C., et al. (2021) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2021.3055830 · Cited by 79
Kim, J.-H., Lee, J., Lee, J., Heo, J., & Kim, J.-Y. (2021) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2020.3039206 · Cited by 78
“Unlike most of the previous works support limited weight bit-precisions such as 1 [17], [18], 1, 2, and 4 bit [16], the Z-PIM can support all weight bit-precisions that vary”
BackgroundResultTwo-Way Transpose Multibit 6T SRAM Computing-in-Memory Macro for Inference-Training AI Edge Chips
Su, J.-W., Si, X., Chou, Y.-C., Chang, T.-W., Huang, W.-H., Tu, Y.-N., Liu, R., Lu, P.-J., Liu, T.-W., Wang, J., et al. (2021) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2021.3108344 · Cited by 78
Bit Parallel 6T SRAM In-memory Computing with Reconfigurable Bit-Precision
Lee, K., Jeong, J., Cheon, S., Choi, W., & Park, J. (2020) · ACM/IEEE Design Automation Conference (DAC) · doi:10.1109/dac18072.2020.9218567 · Cited by 75
A Reliable 8T SRAM for High-Speed Searching and Logic-in-Memory Operations
Chen, J., Zhao, W., Wang, Y., Shu, Y., Jiang, W., & Ha, Y. (2022) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2022.3164756 · Cited by 72
In-memory computing to break the memory wall*
Huang, X., Liu, C., Jiang, Y., & Zhou, P. (2020) · Chinese Physics B · doi:10.1088/1674-1056/ab90e7 · Cited by 79
Su, J.-W., Chou, Y.-C., Liu, R., Liu, T.-W., Lu, P.-J., Wu, P.-C., Chung, Y.-L., Hong, L.-Y., Ren, J.-S., Pan, T., et al. (2022) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2022.3199077 · Cited by 74
“COMPUTING-IN-MEMORY (CIM) combines computing and storage functions within the same macro in order to overcome the memory-wall bottleneck [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19] inherent in the Von Neumann architecture [2], [3], [4], [20], [21], [22], [23]…”
BackgroundMethodA highly stable reliable SRAM cell design for low power applications
Pal, S., Bose, S., Ki, W., & Islam, A. (2019) · Microelectronics Reliability · doi:10.1016/j.microrel.2019.113503 · Cited by 67
A review on SRAM-based computing in-memory: Circuits, functions, and applications
Lin, Z., Tong, Z., Zhang, J., Wang, F., Tian, X., Zhao, Y., Wu, X., Peng, C., Lu, W., Zhao, Q., et al. (2022) · Journal of Semiconductors · doi:10.1088/1674-4926/43/3/031401 · Cited by 72
“used a 10T SRAM to perform a binary line dot product, and the results were reflected on the horizontal source lines[9], conforming to the traditional SRAM storage mode.”
MethodDigital Versus Analog Artificial Intelligence Accelerators: Advances, trends, and emerging designs
Seo, J.-S., Saikia, J., Meng, J., He, W., Suh, H.-S., Anupreetham, A., Liao, Y., Hasssan, A., & Yeo, I. (2022) · IEEE Solid-State Circuits Magazine · doi:10.1109/mssc.2022.3182935 · Cited by 77
“SRAM IMC designs started from single-macro IMC designs, including CONV-SRAM [37], XNOR SRAM [42], Twin-8T SRAM [38], C3SRAM [43], and the 7-nm IMC macro by TSMC [39].”
BackgroundMethodArtificial Intelligence for Metaverse: A Framework
Guo, Y., Yu, T., Wu, J., Wang, Y., Wan, S., Zheng, J., Fang, L., & Dai, Q. (2022) · CAAI Artificial Intelligence Research · doi:10.26599/air.2022.9150004 · Cited by 75
Analysis and Optimization Strategies Toward Reliable and High-Speed 6T Compute SRAM
Chen, J., Zhao, W., Wang, Y., & Ha, Y. (2021) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2021.3054972 · Cited by 56
“When ported to 55nm, the work in [21] will consume 43.8 fJ (Energy ∝ (Technology) 2 [17]), which is 34.7% higher than our work.”
BackgroundMethodInfluentialXuan, L., Un, K.-F., Lam, C., & Martins, R. P. (2022) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2022.3180553 · Cited by 66
“Recently, a plethora of works contributed to the design of energy-efficient DNN accelerators for graphics processing units (GPUs) [2], application-specific integrated circuits (ASICs) [3]–[7], and field-programmable gate arrays (FPGAs) [8]–[10].”
BackgroundAn Embedded nand Flash-Based Compute-In-Memory Array Demonstrated in a Standard Logic Process
Kim, M., Liu, M., Everson, L. R., & Kim, C. H. (2021) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2021.3098671 · Cited by 61
Kim, H., Mu, J., Yu, C., Kim, T. T.-H., & Kim, B. (2023) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2022.3232648 · Cited by 62
“Recently, computing-in-memory (CIM) architectures [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [36], [37], [38] began to receive more attention as a solution for addressing the bottlenecks in the von Neumann architecture by combining the mem…”
BackgroundMethodMu, J., Kim, H., & Kim, B. (2022) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2022.3152653 · Cited by 51
Chougule, S., Chaudhari, B. S., Ghorpade, S. N., & Zennaro, M. (2024) · World Electric Vehicle Journal · doi:10.3390/wevj15020039 · Cited by 58
Sharma, V., Kim, J. E., Kim, H., Lu, L., & Kim, T. T.-H. (2022) · IEEE Journal on Emerging and Selected Topics in Circuits and Systems · doi:10.1109/jetcas.2022.3168571 · Cited by 46
Garofalo, A., Ottavi, G., Conti, F., Karunaratne, G., Boybat, I., Benini, L., & Rossi, D. (2022) · IEEE Journal on Emerging and Selected Topics in Circuits and Systems · doi:10.1109/jetcas.2022.3170152 · Cited by 49
Lin, C.-T., Wang, D., Zhang, B., Chen, G. K., Knag, P. C., Krishnamurthy, R. K., & Seok, M. (2023) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2023.3313519 · Cited by 49
“…test chips demonstrate accessing all rows and performing multiply-and-accumulate (MAC) in one cycle, avoiding the slow row-by-row data access and demonstrating orders of magnitude improvement in energy efficiency and computing throughput [11], [12], [13], [14], [15], [16], [17], [18], [19], [20].”
BackgroundVWA: Hardware Efficient Vectorwise Accelerator for Convolutional Neural Network
Chang, K.-W., & Chang, T. (2019) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2019.2942529 · Cited by 48
“Reference [35] is a computing in memory design that adopts SRAM array for”
MethodA 1–8b Reconfigurable Digital SRAM Compute-in-Memory Macro for Processing Neural Networks
You, H., Li, W., Shang, D., Zhou, Y., & Qiao, S. (2024) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2024.3355944 · Cited by 49
“…have demonstrated the feasibility of CIM operations, such as static random access memory (SRAM) [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], embedded dynamic random access…”
Circuits and Architectures for In-Memory Computing-Based Machine Learning Accelerators
Ankit, A., Chakraborty, I., Agrawal, A., Ali, M., & Roy, K. (2020) · IEEE Micro · doi:10.1109/mm.2020.3025863 · Cited by 46
“Additionally, an inmemory popcount operation can also be embedded in 10T-SRAM cells, where an analog voltage is generated on the sourceline (SL), which represents the output.(4,17) The approach here is to perform a pseudoread operation of a memory row storing an input activation vector.”
Method10T SRAM Computing-in-Memory Macros for Binary and Multibit MAC Operation of DNN Edge Processors
Nguyen, V. T., Kim, J.-S., & Lee, J. (2021) · IEEE Access · doi:10.1109/access.2021.3079425 · Cited by 43
“Compared to [18], [19], which uses multibit MAC operation, the inference accuracy for the MNIST dataset is 2-3% lower; however, the throughput and energy efficiency are 6× and 14× higher, respectively.”
BackgroundMethodResultInfluentialYao, C.-Y., Wu, T., Liang, H.-C., Chen, Y., & Liu, T.-T. (2023) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2022.3224363 · Cited by 46
“This approach is suitable for applications that require only fixed weight and input bit precision [3], [4], [5], [6], [7], [8], [9], [10], [11], [12] but fails to serve the applications demanding the MACs with flexible bit precision.”
MethodWu, P.-C., Su, J.-W., Hong, L.-Y., Ren, J.-S., Chien, C.-H., Chen, H.-Y., Ke, C.-E., Hsiao, H.-M., Li, S.-H., Sheu, S.-S., et al. (2023) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2023.3309966 · Cited by 43
“bottleneck [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15] inherent in the Von Neumann architecture [2], [3], [4], [16], [17], [18], [19].”
Lee, K., Kim, J., & Park, J. (2023) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2023.3334566 · Cited by 45
“Many previous SRAM CIM designs [26], [27], [28], [31] have employed high-resolution ADCs to achieve high accuracy.”
BackgroundMethodInfluentialChen, Z., Wen, Z., Wan, W., Pakala, A., Zou, Y., Wei, W.-C., Li, Z., Chen, Y., & Yang, K. (2024) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2024.3422826 · Cited by 50
“On the other hand, some designs use a current-steering DAC to avoid power-hungry analog buffers [23], [27], [28].”
Wang, J., Bai, Y., Wang, H., Hao, Z., Wang, G., Zhang, K., Zhang, Y., Lv, W., & Zhang, Y. (2022) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2022.3192165 · Cited by 43
“[13] and [38] adopt analog CIM architecture, in which the multiplication and addition results are obtained through bit-line discharging on column.”
BackgroundWu, P.-C., Su, J.-W., Chung, Y.-L., Hong, L.-Y., Ren, J.-S., Chang, F.-C., Wu, Y., Chen, H.-Y., Lin, C.-H., Hsiao, H.-M., et al. (2023) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2023.3343669 · Cited by 42
Chen, Z., Wu, M., Zhou, Y., Li, R., Tan, J., & Ding, D. (2023) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2023.3277517 · Cited by 37
“Biswas and Chandrakasan [21] proposed a 10T SRAM bit cell with two decoupling read ports to solve the problem of read interference.”
BackgroundMethodA 2941-TOPS/W Charge-Domain 10T SRAM Compute-in-Memory for Ternary Neural Network
Cheon, S., Lee, K., & Park, J. (2023) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2023.3241385 · Cited by 47
“Recently, compute-in-memory (CiM) to process DNN operations has been actively researched [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28].”
MethodChen, Y., Yin, G., Zhou, M., Tang, W., Yang, Z., Lee, M., Du, X., Yue, J., Liu, J., Yang, H., et al. (2023) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2023.3266269 · Cited by 38
“Recent works based on SRAM [11], [12], [13], [14], eDRAM [15], [16], and emerging non-volatile memory (NVM) such as RRAM [17], [18], MRAM [19], [20], and FeFET [21], [22] have demonstrated the potential of CiM.”
MethodMC-CIM: Compute-in-Memory With Monte-Carlo Dropouts for Bayesian Edge Intelligence
Shukla, P., Nasrin, S., Darabi, N., Gomes, W., & Trivedi, A. R. (2022) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2022.3224703 · Cited by 40
“memory inference operators; thereby, unlike [8], [9], [10], it does not require digital-to-analog converters (DAC)”
BackgroundA Charge-Sharing based 8T SRAM In-Memory Computing for Edge DNN Acceleration
Lee, K., Cheon, S., Jo, J., Choi, W., & Park, J. (2021) · ACM/IEEE Design Automation Conference (DAC) · doi:10.1109/dac18074.2021.9586103 · Cited by 35
Secure XOR-CIM Engine: Compute-In-Memory SRAM Architecture With Embedded XOR Encryption
Huang, S., Jiang, H., Peng, X., Li, W., & Yu, S. (2021) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2021.3120296 · Cited by 36
Sharma, V., Kim, H., & Kim, T. T.-H. (2022) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2022.3168053 · Cited by 40
Zhang, H., He, S., Lu, X., Guo, X., Wang, S., Du, Y., & Du, L. (2023) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2023.3301814 · Cited by 38
“Restrictions apply. to be consistent across all computing cells and constant throughout the discharging process [17], [18], [27].”
BackgroundA Survey of Computing-in-Memory Processor: From Circuit to Application
Sun, W., Yue, J., He, Y., Huang, Z., Wang, J., Jia, W., Li, Y., Lei, L., Jia, H., & Liu, Y. (2023) · IEEE Open Journal of the Solid-State Circuits Society · doi:10.1109/ojsscs.2023.3328290 · Cited by 49
“Various types of capacitors including MOMCAP [8], [13], MOSCAP [26], and parasitic capacitors [9] are introduced to tradeoff linearity and capacitive density.”
BackgroundSaito, D., Doevenspeck, J., Cosemans, S., Oh, H., Perumkunnil, M., Papistas, I. A., Belmonte, A., Rassoul, N., Delhougne, R., Kar, G. S., et al. (2020) · IEEE Transactions on Electron Devices · doi:10.1109/ted.2020.3025986 · Cited by 34
Paliy, M., Strangio, S., Ruiu, P., Rizzo, T., & Iannaccone, G. (2020) · IEEE Access · doi:10.1109/access.2020.3037017 · Cited by 35
“Analog solutions are also suitable to be implemented with an in-memory circuit architecture [17], [18], avoiding costly memory access.”
BackgroundA 65 nm 73 kb SRAM-Based Computing-In-Memory Macro With Dynamic-Sparsity Controlling
Qiao, X., Song, J., Tang, X., Luo, H., Pan, N., Cui, X., Wang, R., & Wang, Y. (2022) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2022.3162017 · Cited by 31
Datta, G., Kundu, S., Jaiswal, A., & Beerel, P. A. (2022) · Frontiers in Neuroscience · doi:10.3389/fnins.2022.815258 · Cited by 40
Khaddam-Aljameh, R., Francese, P. A., Benini, L., & Eleftheriou, E. S. (2020) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2020.3037871 · Cited by 34
“This is shown in Table III, where recent implementations of analog in-memory MAC-operation accelerators using SRAM combined with capacitors [17], [18], [30], [31] are compared with the presented work.”
BackgroundMethodInfluentialOptical and Electrical Memories for Analog Optical Computing
Kari, S. R., Ríos, C., Jiang, L., Meng, J., Peserico, N., Sorger, V. J., Hu, J., & Youngblood, N. (2023) · IEEE Journal of Selected Topics in Quantum Electronics · doi:10.1109/jstqe.2023.3239918 · Cited by 34
“State-of-the-art SRAM memory using two inverters, which can be in one of two bistable states, has an access latency of 0.21 ns and costs about 5 pJ/bit access [86], [87].”
BackgroundInfluentialIn-Memory Computing: The Next-Generation AI Computing Paradigm
Ma, Y., Du, Y., Du, L., Lin, J., & Wang, Z. (2020) · ACM Great Lakes Symposium on VLSI (GLSVLSI) · doi:10.1145/3386263.3407588 · Cited by 34
“[4] M. Kang, JSSC’18[6] A. Biswas, JSSC’19[10] H. Valavi, JSSC’19[11] J. Wang, JSSC’20[13] X. Si, JSSC’20[12] S. Yin, JSSC’20[”
BackgroundMethodInfluentialA 0.44-μJ/dec, 39.9-μs/dec, Recurrent Attention In-Memory Processor for Keyword Spotting
Dbouk, H., Gonugondla, S. K., Sakr, C., & Shanbhag, N. R. (2020) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2020.3029586 · Cited by 31
SRAM-Based Computing-in-Memory Macro With Fully Parallel One-Step Multibit Computation
Choi, E. J., Choi, I., Jeon, C., Yun, G., Yi, D., Ha, S., Chang, I., & Je, M. (2022) · IEEE Solid-State Circuits Letters · doi:10.1109/lssc.2022.3206416 · Cited by 31
“Our proposed system achieves highest FoM [12] of input precision × weight precision × energy efficiency (scaled to 65-nm technology [8]).”
BackgroundMethodArchitecture of neural processing unit for deep neural networks
Lee, K. J. (2020) · Advances in computers · doi:10.1016/bs.adcom.2020.11.001 · Cited by 43
Process evaluation in FinFET based 7T SRAM cell
Kumar, T. S., & Tripathi, S. L. (2021) · Analog Integrated Circuits and Signal Processing · doi:10.1007/s10470-021-01938-4 · Cited by 28
Towards ADC-Less Compute-In-Memory Accelerators for Energy Efficient Deep Learning
Saxena, U., Chakraborty, I., & Roy, K. (2022) · Design, Automation, and Test in Europe · doi:10.23919/date54114.2022.9774573 · Cited by 36
Guo, R., Yue, Z., Si, X., Li, H., Hu, T., Tang, L., Wang, Y., Sun, H., Liu, L., Chang, M., et al. (2022) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2022.3198413 · Cited by 28
“Since multiply-and-accumulate (MAC) operations are directly integrated into the bit cells of a static random access memory (SRAM) array [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], the weight stationary dataflow [14] matches the CIM paradigm to maximize the reuse of loaded weights.”
BackgroundLin, Z., Zhang, S., Jin, Q., Xia, J., Liu, Y., Yu, K., Zheng, J., Xu, X., Fan, X., Li, K., et al. (2023) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2023.3266651 · Cited by 28
“In this comparison table, works [8], [10], and [14] use an analog calculation method, and the calculation results are affected by the ADC conversion accuracy.”
BackgroundMethodInfluentialChang, L., Li, C., Zhang, Z., Xiao, J., Liu, Q., Zhu, Z., Li, W., Zhu, Z., Yang, S., & Zhou, J. (2021) · Science China Information Sciences · doi:10.1007/s11432-021-3234-0 · Cited by 26
A Ternary Neural Network Computing-in-Memory Processor With 16T1C Bitcell Architecture
Jeong, H., Kim, S., Park, K., Jung, J., & Lee, K. J. (2023) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2023.3265064 · Cited by 26
“Computing-in-Memory (CIM) architecture, which is a promising solution for ultra-low-power IoT devices, has been proposed to replace the conventional computer architecture by processing inside the on-chip memory to achieve high energy efficiency [5], [6], [7], [8], [9].”
BackgroundA Logic Compatible 4T Dual Embedded DRAM Array for In-Memory Computation of Deep Neural Networks
Yoo, T., Kim, H., Chen, Q., Kim, T. T.-H., & Kim, B. (2019) · International Symposium on Low Power Electronics and Design (ISLPED) · doi:10.1109/islped.2019.8824826 · Cited by 32
“SRAM cells in [6-8] have been modified to mitigate the disturb issue.”
BackgroundNasrin, S., Badawi, D., Çetin, A. E., Gomes, W., & Trivedi, A. R. (2021) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2021.3064033 · Cited by 31
Chen, C. C., Dai, Y.-S., & Hong, H.-C. (2024) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2024.3368849 · Cited by 30
“The emerging computing-in-memory (CIM) techniques address these shortcomings by performing the MAC operations directly upon reading the synaptic weights from the memory [5], [6], [7], [8], [9], [10], [11], [12].”
MethodSharma, V., Zhang, X., Dhakad, N. S., & Kim, T. T.-H. (2025) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2025.3547853 · Cited by 32
“Hence, compared with the 10T and 12T used in [4] and [19], the proposed DPU saves 12.5% and 27% transistors for 1b dot-product computation.”
InfluentialRow-wise Accelerator for Vision Transformer
Wang, H., & Chang, T. (2022) · IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS) · doi:10.1109/aicas54282.2022.9869928 · Cited by 26
“However, existing deep learning hardware accelerators [7]–[9] are optimized for CNN based models, which is not suitable for vision transformer models due to the significant difference in the favored architecture structure and parameters.”
BackgroundHe, C., Wang, Z. X., Xiang, F., Dai, Z., He, Y., Yue, J., & Liu, Y. (2023) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2023.3304752 · Cited by 24
XOR-CIM: compute-in-memory SRAM architecture with embedded XOR encryption
Huang, S., Jiang, H., Peng, X., Li, W., & Yu, S. (2020) · doi:10.1145/3400302.3415678 · Cited by 42
“The feasibility of SRAM based CIM accelerators has been proven in silicon prototype chips [10] [11] showing much more competitive energy efficiency compared to the digital ASIC designs (e.”
MethodXu, S., Li, X., Xie, C., Chen, H., Chen, C., & Song, Z. (2021) · Micromachines · doi:10.3390/mi12101183 · Cited by 28
“A recent work experimentally proved functionality (>98% accuracy) in the MNIST dataset [14], using 6-b inputs/outputs.”
BackgroundAn Energy-Efficient and Robust 10T SRAM Based in-Memory Computing Architecture
Srivastava, N., Rajput, A. K., Pattanaik, M., & Kaushal, G. (2023) · International Conference on VLSI Design · doi:10.1109/vlsid57277.2023.00039 · Cited by 24
“SRAM Proposed 10T conv 10T [13] 9T [10]”
BackgroundMethodInfluentialA Practical Design-Space Analysis of Compute-in-Memory With SRAM
Spetalnick, S. D., & Raychowdhury, A. (2022) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2021.3138057 · Cited by 30
“purpose analog add-ons are characteristic of indirect systems [15] and are typified by well-matched metal-oxide-metal (MOM) capacitors [16], [17], shared transistors with improved characteristics [18]–[21], or time-domain computing [22].”
BackgroundAn Efficient and Accurate Memristive Memory for Array-Based Spiking Neural Networks
Das, H., Febbo, R., Tushar, S. N. B., Chakraborty, N. N., Liehr, M., Cady, N. C., & Rose, G. S. (2023) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2023.3301020 · Cited by 30
“A classical example of in-memory computing is the dot product engine (DPE) [6], [7], [8], which can be the backbone of many machine learning algorithms including neuromorphic computing [9], [10] and reservoir computing [11], [12].”
BackgroundA 51.3-TOPS/W, 134.4-GOPS In-Memory Binary Image Filtering in 65-nm CMOS
Bose, S. K., Singla, D., & Basu, A. (2021) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2021.3098539 · Cited by 28
“In general, the SRAM-based realization of near or in-memory computing can be categorized as charge sharing [16], [17], current summation [18], [19], delay-based [20] and xor-based digital [21] implementation.”
Kim, G., Ko, D. H., Kim, T., Lee, S.-H., Jung, M., Lee, Y. K., Lim, S., Jo, M., Eom, T., Shin, H., et al. (2022) · ACS Applied Materials & Interfaces · doi:10.1021/acsami.2c14867 · Cited by 22
Yamaguchi, M., Iwamoto, G., Nishimura, Y., Tamukoh, H., & Morie, T. (2020) · IEEE Access · doi:10.1109/access.2020.3047619 · Cited by 26
“…multi-bit inputs achieved an energy efficiency of up to 10 tera-operations per second per watt (TOPS/W) [24], that for the BinaryConnect model achieved up to about 50 TOPS/W [25], and those for binarized networks achieved up to 700 TOPS/W with the aid of an analog computing circuit approach [26].”
MethodToward Real-Time, At-Home Patient Health Monitoring Using Reservoir Computing CMOS IC
Chandrasekaran, S. T., Bhanushali, S. P., Banerjee, I., & Sanyal, A. (2021) · IEEE Journal on Emerging and Selected Topics in Circuits and Systems · doi:10.1109/jetcas.2021.3128587 · Cited by 22
Chen, Y., Yin, G., Tan, Z., Lee, M., Yang, Z., Liu, Y., Yang, H., Ma, K., & Li, X. (2022) · ACM/IEEE Design Automation Conference (DAC) · doi:10.1145/3489517.3530576 · Cited by 31
“Figure 4: Proposed 1T/cell ROM-based CiM cell in (a), and some existing SRAM-CiM cells in (b-f) [3][4][19][20][21].”
BackgroundResultThe 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
“Biswas et al utilized 1-b multiplication-and-average operation in analog domain right next to the 10T SRAM cell array to accelerate binary-weight network [109] and Yin et al proposed SRAM macro that computes ternary-XNOR-and-accumulate operation.”
MethodiMCU: A 28-nm Digital In-Memory Computing-Based Microcontroller Unit for TinyML
Lin, C.-T., Huang, P. X., Oh, J., Wang, D., & Seok, M. (2024) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2024.3362274 · Cited by 22
“SRAM-based in-memory computing (IMC) has been proposed for improving energy efficiency and throughput in vector–matrix multiplication (VMM) [8], [9], [10], [11], [12], [13], [14], [15].”
Multi-Bit Compute-In Memory Architecture Using a C-2C Ladder Network
Kushwaha, D., Abotula, J. K., Kohli, R., Mishra, J., Dasgupta, S., & Bulusu, A. (2023) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2023.3329261 · Cited by 21
“Thus, many recent works [4], [5], [6], [7], [8], [9], [10], [11], [12] approached compute in memory (CIM) to alleviate the Von Neumann bottleneck and perform the fundamental operation of CNN: MAC (multiply and accumulate) in the memory itself, which eliminates the huge power consumption for data…”
BackgroundAn 11T SRAM Cell for the Dual-Direction In-Array Logic/CAM Operations
Wei, F., Cui, X., Zhang, S., & Zhang, X. (2023) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2023.3337119 · Cited by 24
“The in-memory multiply accumulation (MAC) computations are discussion in [1], [2], [3], [4], [5], [6], [7], [8], [9], [10].”
Metadata
- Authors
- Biswas, A., & Chandrakasan, A. P.
- Affiliations
- Texas Instruments (United States) (US)
- Massachusetts Institute of Technology (US)
- Venue
- IEEE Journal of Solid-State Circuits
- Published
- 2019-01
- Volume
- 54
- Issue
- 1
- Pages
- 217-230
- ISSN
- 0018-9200 (ISSN-L)
- 1558-173X
- OpenAlex
- W2904299207
- Cited by
- 414
- References
- 29
- Citations by year
- 2026: 36 · 2025: 57 · 2024: 75 · 2023: 75 · 2022: 74 · 2021: 59 · 2020: 24 · 2019: 13
- FWCI
- 9.17
- Type
- journal article
- Language
- en
- Open access
- green
- hdl.handle.net
- Licence
- Locations
- IEEE Journal of Solid-State Circuits (publishedVersion)
- DSpace@MIT (Massachusetts Institute of Technology) (submittedVersion, cc-by-nc-sa)
- Funders
- Intel Corporation
- Topics
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Advanced Memory and Neural Computing
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Low-power high-performance VLSI design
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Analog and Mixed-Signal Circuit Design
- Keywords
- SRAM, in-memory computing, dot product computation, energy efficiency, low-power neural network hardware, analog-to-digital converter, MNIST classification
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.