The First Embedded 14nm FeFinFET NVM: 2T1CFE Array as Electrical Synapses and Activations for High-performance and Low-power Inference Accelerators

E. R. Hsieh; W. L. Tsai; Y. L. Lin; C. H. Liu; S. S. Chung; Y. T. Tang; J. R. Su; T. P. Chen; S. A. Huang; T. J. Chen; O. Cheng

The First Embedded 14nm FeFinFET NVM: 2T1C_FE Array as Electrical Synapses and Activations for High-performance and Low-power Inference Accelerators

E. R. Hsieh^*, W. L. Tsai, Y. L. Lin, C. H. Liu, S. S. Chung, Y. T. Tang, J. R. Su, T. P. Chen, S. A. Huang, T. J. Chen, O. Cheng

^*Corresponding author for this work

Department of Mechatronic Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

3 Citations (Scopus)

Abstract

We present the 1^st embedded 14nm FeFinFET NVMs, 2T1CFE. Compared to 1T1CFE array, 2T1CFE one shows 25x↑of the SN ratio, 10⁴x↓of the PGM power, and 10³x↓of READ power. The gradual tuning of conductance for 2T1CFE has been tuned linearly and symmetrically with 3x10⁴x of the window, where 8 separated conductance states can be stored with crystal gaps in between. Continual endurance cycles for 8 states are >10⁷ times. The retention of 8 states has been baked in 85°C for >1 month. The 8 states have also passed the PGM disturbance test. Finally, this 2T1CFE array is as electrical synapses in a very deep neural network (DNN) with 23.8 million parameters for practice, and most importantly, the 2T1CFE is also as an activation, Rectified Linear Unit (ReLU), for the first time. Compared to the 5T-CMOS ReLU, the 2T1CFE ReLU shows 70%↑of inference accuracy. Overall, 2T1CFE shows strong potential as key components in high-performance low-power inference-accelerators.

Original language	English
Title of host publication	2021 Symposium on VLSI Technology, VLSI Technology 2021
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)	9784863487802
Publication status	Published - 2021
Event	41st Symposium on VLSI Technology, VLSI Technology 2021 - Virtual, Online, Japan Duration: 2021 Jun 13 → 2021 Jun 19

Publication series

Name	Digest of Technical Papers - Symposium on VLSI Technology
Volume	2021-June
ISSN (Print)	0743-1562

Conference

Conference	41st Symposium on VLSI Technology, VLSI Technology 2021
Country/Territory	Japan
City	Virtual, Online
Period	2021/06/13 → 2021/06/19

ASJC Scopus subject areas

Electrical and Electronic Engineering

Cite this

Hsieh, E. R., Tsai, W. L., Lin, Y. L., Liu, C. H., Chung, S. S., Tang, Y. T., Su, J. R., Chen, T. P., Huang, S. A., Chen, T. J., & Cheng, O. (2021). The First Embedded 14nm FeFinFET NVM: 2T1C_FE Array as Electrical Synapses and Activations for High-performance and Low-power Inference Accelerators. In 2021 Symposium on VLSI Technology, VLSI Technology 2021 (Digest of Technical Papers - Symposium on VLSI Technology; Vol. 2021-June). Institute of Electrical and Electronics Engineers Inc..

The First Embedded 14nm FeFinFET NVM: 2T1C_FE Array as Electrical Synapses and Activations for High-performance and Low-power Inference Accelerators. / Hsieh, E. R.; Tsai, W. L.; Lin, Y. L. et al.
2021 Symposium on VLSI Technology, VLSI Technology 2021. Institute of Electrical and Electronics Engineers Inc., 2021. (Digest of Technical Papers - Symposium on VLSI Technology; Vol. 2021-June).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Hsieh, ER, Tsai, WL, Lin, YL, Liu, CH, Chung, SS, Tang, YT, Su, JR, Chen, TP, Huang, SA, Chen, TJ & Cheng, O 2021, The First Embedded 14nm FeFinFET NVM: 2T1C_FE Array as Electrical Synapses and Activations for High-performance and Low-power Inference Accelerators. in 2021 Symposium on VLSI Technology, VLSI Technology 2021. Digest of Technical Papers - Symposium on VLSI Technology, vol. 2021-June, Institute of Electrical and Electronics Engineers Inc., 41st Symposium on VLSI Technology, VLSI Technology 2021, Virtual, Online, Japan, 2021/06/13.

Hsieh ER, Tsai WL, Lin YL, Liu CH, Chung SS, Tang YT et al. The First Embedded 14nm FeFinFET NVM: 2T1C_FE Array as Electrical Synapses and Activations for High-performance and Low-power Inference Accelerators. In 2021 Symposium on VLSI Technology, VLSI Technology 2021. Institute of Electrical and Electronics Engineers Inc. 2021. (Digest of Technical Papers - Symposium on VLSI Technology).

Hsieh, E. R. ; Tsai, W. L. ; Lin, Y. L. et al. / The First Embedded 14nm FeFinFET NVM : 2T1C_FE Array as Electrical Synapses and Activations for High-performance and Low-power Inference Accelerators. 2021 Symposium on VLSI Technology, VLSI Technology 2021. Institute of Electrical and Electronics Engineers Inc., 2021. (Digest of Technical Papers - Symposium on VLSI Technology).

@inproceedings{7d51bb278cf144888c363c86261a932c,

title = "The First Embedded 14nm FeFinFET NVM: 2T1CFE Array as Electrical Synapses and Activations for High-performance and Low-power Inference Accelerators",

abstract = "We present the 1st embedded 14nm FeFinFET NVMs, 2T1CFE. Compared to 1T1CFE array, 2T1CFE one shows 25x↑of the SN ratio, 104x↓of the PGM power, and 103x↓of READ power. The gradual tuning of conductance for 2T1CFE has been tuned linearly and symmetrically with 3x104x of the window, where 8 separated conductance states can be stored with crystal gaps in between. Continual endurance cycles for 8 states are >107 times. The retention of 8 states has been baked in 85°C for >1 month. The 8 states have also passed the PGM disturbance test. Finally, this 2T1CFE array is as electrical synapses in a very deep neural network (DNN) with 23.8 million parameters for practice, and most importantly, the 2T1CFE is also as an activation, Rectified Linear Unit (ReLU), for the first time. Compared to the 5T-CMOS ReLU, the 2T1CFE ReLU shows 70%↑of inference accuracy. Overall, 2T1CFE shows strong potential as key components in high-performance low-power inference-accelerators.",

author = "Hsieh, {E. R.} and Tsai, {W. L.} and Lin, {Y. L.} and Liu, {C. H.} and Chung, {S. S.} and Tang, {Y. T.} and Su, {J. R.} and Chen, {T. P.} and Huang, {S. A.} and Chen, {T. J.} and O. Cheng",

note = "Publisher Copyright: {\textcopyright} 2021 JSAP; 41st Symposium on VLSI Technology, VLSI Technology 2021 ; Conference date: 13-06-2021 Through 19-06-2021",

year = "2021",

language = "English",

series = "Digest of Technical Papers - Symposium on VLSI Technology",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

booktitle = "2021 Symposium on VLSI Technology, VLSI Technology 2021",

}

TY - GEN

T1 - The First Embedded 14nm FeFinFET NVM

T2 - 41st Symposium on VLSI Technology, VLSI Technology 2021

AU - Hsieh, E. R.

AU - Tsai, W. L.

AU - Lin, Y. L.

AU - Liu, C. H.

AU - Chung, S. S.

AU - Tang, Y. T.

AU - Su, J. R.

AU - Chen, T. P.

AU - Huang, S. A.

AU - Chen, T. J.

AU - Cheng, O.

PY - 2021

Y1 - 2021

N2 - We present the 1st embedded 14nm FeFinFET NVMs, 2T1CFE. Compared to 1T1CFE array, 2T1CFE one shows 25x↑of the SN ratio, 104x↓of the PGM power, and 103x↓of READ power. The gradual tuning of conductance for 2T1CFE has been tuned linearly and symmetrically with 3x104x of the window, where 8 separated conductance states can be stored with crystal gaps in between. Continual endurance cycles for 8 states are >107 times. The retention of 8 states has been baked in 85°C for >1 month. The 8 states have also passed the PGM disturbance test. Finally, this 2T1CFE array is as electrical synapses in a very deep neural network (DNN) with 23.8 million parameters for practice, and most importantly, the 2T1CFE is also as an activation, Rectified Linear Unit (ReLU), for the first time. Compared to the 5T-CMOS ReLU, the 2T1CFE ReLU shows 70%↑of inference accuracy. Overall, 2T1CFE shows strong potential as key components in high-performance low-power inference-accelerators.

AB - We present the 1st embedded 14nm FeFinFET NVMs, 2T1CFE. Compared to 1T1CFE array, 2T1CFE one shows 25x↑of the SN ratio, 104x↓of the PGM power, and 103x↓of READ power. The gradual tuning of conductance for 2T1CFE has been tuned linearly and symmetrically with 3x104x of the window, where 8 separated conductance states can be stored with crystal gaps in between. Continual endurance cycles for 8 states are >107 times. The retention of 8 states has been baked in 85°C for >1 month. The 8 states have also passed the PGM disturbance test. Finally, this 2T1CFE array is as electrical synapses in a very deep neural network (DNN) with 23.8 million parameters for practice, and most importantly, the 2T1CFE is also as an activation, Rectified Linear Unit (ReLU), for the first time. Compared to the 5T-CMOS ReLU, the 2T1CFE ReLU shows 70%↑of inference accuracy. Overall, 2T1CFE shows strong potential as key components in high-performance low-power inference-accelerators.

UR - http://www.scopus.com/inward/record.url?scp=85125474382&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85125474382&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:85125474382

T3 - Digest of Technical Papers - Symposium on VLSI Technology

BT - 2021 Symposium on VLSI Technology, VLSI Technology 2021

PB - Institute of Electrical and Electronics Engineers Inc.

Y2 - 13 June 2021 through 19 June 2021

ER -