An Accurate, Error-Tolerant, and Energy-Efficient Neural Network Inference Engine Based on SONOS Analog Memory
- 4 January 2022
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. 69 (4) , 1480-1493
Abstract
We demonstrate SONOS (silicon-oxide-nitride-oxide-silicon) analog memory arrays that are optimized for neural network inference. The devices are fabricated in a 40nm process and operated in the subthreshold regime for in-memory matrix multiplication. Subthreshold operation enables low conductances to be implemented with low error, which matches the typical weight distribution of neural networks, which is heavily skewed toward near-zero values. This leads to high accuracy in the presence of programming errors and process variations. We simulate the end-to-end neural network inference accuracy, accounting for the measured programming error, read noise, and retention loss in a fabricated SONOS array. Evaluated on the ImageNet dataset using ResNet50, the accuracy using a SONOS system is within 2.16% of floating-point accuracy without any retraining. The unique error properties and high On/Off ratio of the SONOS device allow scaling to large arrays without bit slicing, and enable an inference architecture that achieves 20 TOPS/W on ResNet50, a $> 10\times $ gain in energy efficiency over state-of-the-art digital and analog inference accelerators.Keywords
Funding Information
- Laboratory Directed Research and Development Program at Sandia National Laboratories
- Defense Threat Reduction Agency (HDTRA1-17-1-0038)
This publication has 56 references indexed in Scilit:
- PRIME: A Novel Processing-in-Memory Architecture for Neural Network Computation in ReRAM-Based Main Memory2016
- Dot-product engine for neuromorphic computing2016
- Resistive switching memories based on metal oxides: mechanisms, reliability and scalingSemiconductor Science and Technology, 2016
- Memristive Boltzmann machine: A hardware accelerator for combinatorial optimization and deep learning2016
- Data retention in MLC NAND flash memory: Characterization, optimization, and recovery2015
- Access devices for 3D crosspoint memoryJournal of Vacuum Science & Technology B, 2014
- The missing memristor foundNature, 2008
- Performance and Reliability Features of Advanced Nonvolatile Memories Based on Discrete Traps (Silicon Nanocrystals, SONOS)IEEE Transactions on Device and Materials Reliability, 2004
- New second generation current conveyor with reduced parasitic resistance and bandpass filter applicationIEEE Transactions on Circuits and Systems I: Regular Papers, 2001
- A unified model for the flicker noise in metal-oxide-semiconductor field-effect transistorsIEEE Transactions on Electron Devices, 1990