EUVIMEDEUVIMED entdecken
Crossref

AgeBalance: Low-Cost Lifetime Extension for SRAM-Based PIM Accelerators

Ning Lin, Shaocong Wang, Yaping Zhao, Yangu He, Songqi Wang, Kwunhang Wong, Rongliang Fu, Wenxing Li, Tsung-Yi Ho, Dashan Shang, Xiaojuan Qi, Xiaoming Chen und 1 weitere

IEEE Transactions on Computers · 2026-08 · 75 (8) · 2870-2884

DOI: 10.1109/tc.2026.3695104 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Computers, 2026-08. Verfasst von Ning Lin und 12 weiteren Autor:innen. Die Arbeit nennt 35 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Hong Kong Research Grants Council under the General Research Fund Scheme, Theme-based Research Scheme.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Although processing-in-memory (PIM) techniques have widely been used for deep neural networks (DNNs) acceleration, the inference performance of aged PIM-based accelerators remains to be investigated. This paper makes the first attempt to study Hot Carrier Injection (HCI) and Negative Bias Temperature Instability (NBTI) aging impacts on SRAM-based DNN accelerators, which provides a novel and unified framework, termed AgeBalance for aging detection, analysis and mitigation. First, we discuss a convenient aging detection scheme. Then, we benchmark the inference accuracy drops of DNNs running on aged SRAM-based PIM accelerators. Finally, we propose a low-cost anti-aging training method without incurring additional hardware overhead on SRAM-based DNN accelerators. Extensive experimental results on MNIST, CIFAR10 and AG News datasets show that aging can cause the inference accuracy of shallow or deep DNNs to drop to about 10%, close to random guessing. The aging mitigation scheme proposed in this paper can largely restore the accuracy to the original. Moreover, the SRAM write overhead of our method is much reduced thanks to a score-based training approach, leading to a reduction of 5 $\times$ × to 10 $\times$ × writing energy compared to the traditional training method.

Förderung

  • National Natural Science Foundation of China
  • Hong Kong Research Grants Council under the General Research Fund Scheme
  • Theme-based Research Scheme
  • Strategic Topics
  • JC STEM Lab of Robotics for Soft Materials, funded by The Hong Kong Jockey Club Charities Trust

Zitieren

APA

Lin, N., Wang, S., Zhao, Y., He, Y., Wang, S., Wong, K., Fu, R., Li, W., Ho, T., Shang, D., Qi, X., Chen, X., & Wang, Z. (2026). AgeBalance: Low-Cost Lifetime Extension for SRAM-Based PIM Accelerators. IEEE Transactions on Computers, 75(8), 2870-2884. https://doi.org/10.1109/tc.2026.3695104

Vancouver

Lin N, Wang S, Zhao Y, He Y, Wang S, Wong K, et al. AgeBalance: Low-Cost Lifetime Extension for SRAM-Based PIM Accelerators. IEEE Transactions on Computers. 2026;75(8):2870-2884. doi:10.1109/tc.2026.3695104

BibTeX
@article{Lin2026,
  author = {Lin, Ning and Wang, Shaocong and Zhao, Yaping and He, Yangu and Wang, Songqi and Wong, Kwunhang and Fu, Rongliang and Li, Wenxing and Ho, Tsung-Yi and Shang, Dashan and Qi, Xiaojuan and Chen, Xiaoming and Wang, Zhongrui},
  title = {AgeBalance: Low-Cost Lifetime Extension for SRAM-Based PIM Accelerators},
  journal = {IEEE Transactions on Computers},
  year = {2026},
  volume = {75},
  number = {8},
  pages = {2870--2884},
  doi = {10.1109/tc.2026.3695104}
}
Kostenlos starten →