The DOI matches its Crossref record; 39 of the 39 references it cites were found in a registry or an index; 1 of the 17 sources its claims cite were checked for support.
A computing-in-memory macro based on three-dimensional resistive random-access memory
Open accessAbstract
Abstract Non-volatile computing-in-memory macros that are based on two-dimensional arrays of memristors are of use in the development of artificial intelligence edge devices. Scaling such systems to three-dimensional arrays could provide higher parallelism, capacity and density for the necessary vector–matrix multiplication operations. However, scaling to three dimensions is challenging due to manufacturing and device variability issues. Here we report a two-kilobit non-volatile computing-in-memory macro that is based on a three-dimensional vertical resistive random-access memory fabricated using a 55 nm complementary metal–oxide–semiconductor process. Our macro can perform 3D vector–matrix multiplication operations with an energy efficiency of 8.32 tera-operations per second per watt when the input, weight and output data are 8, 9 and 22 bits, respectively, and the bit density is 58.2 bit µm –2 . We show that the macro offers more accurate brain MRI edge detection and improved inference accuracy on the CIFAR-10 dataset than conventional methods.
Overview
CoChat Summary
## TL;DR
This paper reports a 2-kb three-dimensional vertical RRAM computing-in-memory macro fabricated in 55 nm CMOS, paired with circuit schemes to reduce conductance-drift error and latency. The macro achieves up to 62.11 TOPS/W in 1bIN-2bW mode and improves edge-detection and classification accuracy versus conventional approaches.
## Key findings
- The authors report a 2-kb 3D VRRAM-based nvCIM macro fabricated using a 55 nm CMOS process, with a bit density of 58.2 bit µm^-2.
- The macro achieves energy efficiency of 62.11 TOPS/W at 1bIN-2bW, 29.94 TOPS/W at 4bIN-5bW, and 8.32 TOPS/W at 8bIN-9bW.
- The proposed ADINWM + CADS approach reduces error from conductance drift; for 1bIN-2bW on MNIST, accuracy improves from 93.89% to 94.70% under PWIVMM comparison, and on CIFAR-10 the ADINWM scheme reaches 90.54%.
- For the 4bIN-5bW and 8bIN-9bW configurations, the hardware results lose only 0.81% and 0.84% inference accuracy, respectively, relative to pure software simulation on MNIST with the self-defined CNN.
## Methods
- Design: A 3D VRRAM-based nvCIM macro combines a multilevel self-selective (MLSS) 3D VRRAM array with ADINWM, including CADS for current shaping, plus an analog multiplier (AM) and direct small-current converter (DSCC) to support low-latency computation.
- Sample / data: The hardware platform is built around a 2-kb 3D VRRAM array with peripheral circuits fabricated in 55 nm CMOS; evaluation includes 3D brain MRI edge detection, MNIST, and CIFAR-10.
- Main analysis: The paper compares the proposed ADINWM scheme against conventional PWIVMM, using FPGA-controlled hardware testing, software simulation, and inference-accuracy/energy-efficiency measurements across 1bIN-2bW, 4bIN-5bW, and 8bIN-9bW configurations.
## Limitations
- The paper states that the platform is limited by current technology: inference on CIFAR-10 is done by repeatedly writing CNN weights to 3D VRRAM, but too many write operations may break down the device.
- The authors note that the current work is a proof-of-concept and that larger 3D VRRAMs or multiple arrays are needed for larger CNNs.
- Evident from the text: the reported chip is a 2-kb macro and the eight-layer array is 8 × 8 × 32, so the demonstrated system is small-scale and not yet a full large-network deployment.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
1 Valid · 6 Not valid · 7 Not checked · 53 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“Multibit weights are composed of low-bit weights 33 through the cooperation of peripheral circuits.”
What the sources say
- Supportedcites [33] Irem, B. et al. Neuromorphic computing with multi-memristive synapses. Nat. Commun. 9, 2514 (2018). source
The citation follows “low-bit weights,” so the attributed point is that a higher-resolution weight is formed from multiple constituent devices. The Results section (“The multi-memristive synapse”) says the synaptic weight is represented by the combined conductance of N devices, increasing its resolution. (LLM, conf=0.89)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The SSC is realized by using a bilayer HfO2/TaO x device.”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Our 3D nvCIM macro achieves computing precision with a range of input-, weight- and output-bit levels, and offers improved inference accuracies compared with conventional approaches when tested using the Modified National Institute of Standards and Technology (MNIST) and Canadian Institute For Advanced Research (CIFAR-10) datasets.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Our macro can perform 3D vector-matrix multiplication operations with an energy efficiency of 8.32 tera-operations per second per watt when the input, weight and output data are 8, 9 and 22 bits, respectively, and the bit density is 58.2 bit µ m -2 .”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The 3D VRRAM only contains WLs and BLs because of using a self-selective cell (SSC) to avoid the introduction of planar electrodes and select line (SL) in conventional 3D VRRAMs.”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
35 more claims
Create a free account to see all 67 claims and to verify the 3 references still open.
- It may state two separate claims as one. 22
- The quote only partly states this claim. 1
- Probably not a substantial claim. 7
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
LeCun, Y., Bengio, Y., & Hinton, G. (2015) · Nature · doi:10.1038/nature14539 · Cited by 84,126
3D Convolutional Neural Networks for Human Action Recognition
Ji, S., Xu, W., Yang, M., & Yu, K. (2013) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2012.59 · Cited by 6,235
Efficient 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 2×
“T he use of convolutional neural networks (CNNs) has led to improvements in computer vision, natural language processing and speech recognition. tremendous These methods do, however, have computational and storage demands 1–3 .”
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
Fully hardware-implemented memristor convolutional neural network
Yao, P., Wu, H., Gao, B., Tang, J., Zhang, Q., Zhang, W., Yang, J. J., & Qian, H. (2020) · Nature · doi:10.1038/s41586-020-1942-4 · Cited by 2,179 · cited 1×
“For the 8bIN–9bW operation (2-bit cell), S1, S2 and ) Yao et al. 13 Liu et al. 16 Fig.”
MethodSelf-citationChi, P., Li, S., Xu, C., Zhang, T., Zhao, J., Liu, Y., Wang, Y., & Xie, Y. (2016) · ACM SIGARCH Computer Architecture News · doi:10.1145/3007787.3001140 · Cited by 1,231
Self-citationEIE: Efficient Inference Engine on Compressed Deep Neural Network
Han, S., Liu, X., Mao, H., Pu, J., Pedram, A., Horowitz, M. A., & Dally, W. J. (2016) · 2016 ACM/IEEE 43rd Annual International Symposium on Computer Architecture (ISCA) · doi:10.1109/isca.2016.30 · Cited by 1,077
Neuromorphic computing with multi-memristive synapses
Boybat, I., Le Gallo, M., Nandakumar, S. R., Moraitis, T., Parnell, T., Tuma, T., Rajendran, B., Leblebici, Y., Sebastian, A., & Eleftheriou, E. (2018) · Nature Communications · doi:10.1038/s41467-018-04933-y · Cited by 899 · cited 1×
“Multibit weights are composed of low-bit weights 33 through the cooperation of peripheral circuits.”
BackgroundThree-dimensional integration of nanotechnologies for computing and data storage on a single chip
Shulaker, M. M., Hills, G., Park, R. S., Howe, R. T., Saraswat, K., Wong, H.-S. P., & Mitra, S. (2017) · Nature · doi:10.1038/nature22994 · Cited by 866
Cited by
Flexible brain–computer interfaces
Tang, X., Shen, H., Zhao, S., Li, N., & Liu, J. (2023) · Nature Electronics · doi:10.1038/s41928-022-00913-9 · Cited by 373
Le 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
Programming memristor arrays with arbitrarily high precision for analog computing
Song, W., Rao, M., Li, Y., Li, C., Zhuo, Y., Cai, F., Wu, M., Yin, W., Li, Z., Wei, Q., et al. (2024) · Science · doi:10.1126/science.adi9405 · Cited by 193
Chen, S., Zhang, T., Tappertzhofen, S., Yang, Y., & Valov, I. (2023) · Advanced Materials · doi:10.1002/adma.202301924 · Cited by 177
Self-citationA 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 198
Self-citation
Metadata
- Authors
- Huo, Q., Yang, Y., Wang, Y., Lei, D., Fu, X., Ren, Q., Xu, X., Luo, Q., Xing, G., Chen, C., Si, X., Wu, H., Yuan, Y., Li, Q., Li, X., Wang, X., Chang, M.-F., Zhang, F., & Liu, M.
- Affiliations
- Chinese Academy of Sciences (CN)
- Beijing Institute of Technology (CN)
- Guangdong University of Technology (CN)
- Xiamen University of Technology (CN)
- University of Electronic Science and Technology of China (CN)
- National Tsing Hua University (TW)
- Corresponding author
- Wang, Xinghua
- Zhang, Feng
- ORCID
- Venue
- Nature Electronics
- Published
- 2022-07-26
- Volume
- 5
- Issue
- 7
- Pages
- 469-477
- ISSN
- 2520-1131 (ISSN-L)
- OpenAlex
- W4288068957
- Semantic Scholar
- 88b31544110c39d64621e66cb7955365bf14e368
- Cited by
- 159
- References
- 38
- Citations by year
- 2026: 36 · 2025: 46 · 2024: 33 · 2023: 38 · 2022: 6
- FWCI
- 8.38
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by
- creativecommons.org/licenses/by/4.0 (tdm)
- creativecommons.org/licenses/by/4.0 (vor)
- Locations
- Nature Electronics (publishedVersion, cc-by)
- Funders
- CAS | Institute of Microelectronics of the Chinese Academy of Sciences (2018YFB0407500,)
- 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 › Ferroelectric and Negative Capacitance Devices
- Keywords
- in-memory computing, resistive random access memory, memristors, non-volatile memory, matrix-vector multiplication, energy efficiency, edge inference, CIFAR-10 classification, device variability
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.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
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
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: A computing-in-memory macro based on three-dimensional resistive random-access memory. https://app.cochat.ai/papers/10.1038/s41928-022-00795-x
- The paper
- Huo, Q., Yang, Y., Wang, Y., Lei, D., Fu, X., Ren, Q., Xu, X., Luo, Q., Xing, G., Chen, C., Si, X., Wu, H., Yuan, Y., Li, Q., Li, X., Wang, X., Chang, M.-F., Zhang, F., & Liu, M. (2022). A computing-in-memory macro based on three-dimensional resistive random-access memory. Nature Electronics. https://doi.org/10.1038/s41928-022-00795-x
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.