EUVIMEDEUVIMED entdecken
Crossref

MACC-SRAM: A Multistep Accumulation Capacitor-Coupling In-Memory Computing SRAM Macro for Deep Convolutional Neural Networks

Bo Zhang, Jyotishman Saikia, Jian Meng, Dewei Wang, Soonwan Kwon, Sungmeen Myung, Hyunsoo Kim, Sang Joon Kim, Jae-Sun Seo, Mingoo Seok

IEEE Journal of Solid-State Circuits · 2024-06 · 59 (6) · 1938-1949

DOI: 10.1109/jssc.2023.3332017 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Journal of Solid-State Circuits, 2024-06. Verfasst von Bo Zhang und 9 weiteren Autor:innen. Die Arbeit wurde bisher 50-mal zitiert (Crossref) und nennt 37 Literaturangaben. Gefördert von: Samsung Electronics and CoCoSys, one of seven centers in JUMP 2.0, an Semiconductor Research Corporation (SRC) program sponsored by Defense Advanced Research P….

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This article presents multistep accumulation capacitor coupling static random-access memory (MACC-SRAM), capacitor-based in-memory computing (IMC) SRAM macro for 4-b deep convolutional neural network (DNN) inference. The macro can simultaneously activate all its 128 $\times $ 128 custom 9T1C bitcells to perform the vector-matrix multiplication (VMM). MACC-SRAM also integrates 128 stepwise-charging and discharging input drivers (SCD-IDRs) to efficiently convert the digital codes of the input activations into analog voltages in a 2-b serial fashion. As a result, it can save up to 66% of the capacitor-driving energy. Also, the macro adopts an adder-first architecture to reduce the analog-to-digital (A/D) conversion overhead for the analog-mixed-signal (AMS) computation. The partial sums of the four adjacent rows, representing different bit positions in the 4-b weights, are first accumulated with an analog switched-capacitor adder and then converted to digital codes by a 6-bit successive approximation register (SAR) analog-to-digital converter (ADC). Compared with the ADC-first architecture, where partial sums of each row are first converted to digital codes and then accumulated in the digital domain, the adder-first architecture can save 60.7% of the area and 66.7% of the energy consumption of the A/D conversion. Moreover, the co-optimization of the DNN model by increasing the sparsity further reduces 39.4% of the capacitor-driving energy with a neglectable 0.4% DNN accuracy loss. We prototyped the macro in 28 nm technology, and the measurement shows an energy efficiency of 163 tera-operations per second per watt (TOPS/W) and a throughput of 211 giga-operations per second (GOPS) for a 4-b/4-b DNN model under 0.9 V supply, among the highest of the recent IMC SRAMs.

Förderung

  • Samsung Electronics and CoCoSys, one of seven centers in JUMP 2.0, an Semiconductor Research Corporation (SRC) program sponsored by Defense Advanced Research P…

Zitieren

APA

Zhang, B., Saikia, J., Meng, J., Wang, D., Kwon, S., Myung, S., Kim, H., Kim, S. J., Seo, J., & Seok, M. (2024). MACC-SRAM: A Multistep Accumulation Capacitor-Coupling In-Memory Computing SRAM Macro for Deep Convolutional Neural Networks. IEEE Journal of Solid-State Circuits, 59(6), 1938-1949. https://doi.org/10.1109/jssc.2023.3332017

Vancouver

Zhang B, Saikia J, Meng J, Wang D, Kwon S, Myung S, et al. MACC-SRAM: A Multistep Accumulation Capacitor-Coupling In-Memory Computing SRAM Macro for Deep Convolutional Neural Networks. IEEE Journal of Solid-State Circuits. 2024;59(6):1938-1949. doi:10.1109/jssc.2023.3332017

BibTeX
@article{Zhang2024,
  author = {Zhang, Bo and Saikia, Jyotishman and Meng, Jian and Wang, Dewei and Kwon, Soonwan and Myung, Sungmeen and Kim, Hyunsoo and Kim, Sang Joon and Seo, Jae-Sun and Seok, Mingoo},
  title = {MACC-SRAM: A Multistep Accumulation Capacitor-Coupling In-Memory Computing SRAM Macro for Deep Convolutional Neural Networks},
  journal = {IEEE Journal of Solid-State Circuits},
  year = {2024},
  volume = {59},
  number = {6},
  pages = {1938--1949},
  doi = {10.1109/jssc.2023.3332017}
}
Kostenlos starten →