Analog compute in memory design reduces noise for accurate AI processing
NOVA-CIM: Noise- and Correlation-Tolerant Stochastic Interfaces for Analog Compute-in-Memory
Hardware Architecture
Summary
Computers that do calculations directly in memory can save energy but often struggle with errors caused by noisy signals. The researchers developed a new method called NOVA-CIM that changes how signals are read, using simple yes/no comparisons instead of detailed measurements. This approach averages out random noise and relies less on precise hardware, making the system more reliable. Their tests show that this design keeps accuracy high even when signals are noisy or hardware is reused.
analog compute-in-memoryADCread noisestochastic sensingprobability-domain readouttemporal averagingspatial correlationViT-Baseinference accuracyroot-mean-square error
Authors
Jiachen Ren, Wenshuai Yao, Haobo Liu, Xincheng Feng, Chenxi Hu, Zhengwu Liu, Kechao Tang, Wenyong Zhou, Ngai Wong
Abstract
Analog compute-in-memory (CIM) enables energy-efficient model acceleration, but its reliance on ADC-based readout, which directly quantizes noisy column currents, makes inference accuracy highly sensitive to analog read noise, active-row scaling, and ADC precision. In this paper, we present NOVA-CIM, a noise- and correlation-tolerant stochastic interface for analog CIM by replacing multi-bit ADC readout with random-reference 1-bit sensing and reconstructing results through lightweight counting. By converting column currents into comparison probabilities, this probability-domain readout averages zero-mean dynamic read noise over stochastic samples while reducing dependence on high-resolution ADCs. We provide a unified robustness analysis showing that dynamic read noise is suppressed through temporal averaging and that spatial input-bitstream correlation increases instantaneous current variance rather than introducing first-order MAC bias. MAC-level experiments and end-to-end evaluation on ViT-Base validate the analysis: under read noise, Top-1 accuracy remains 84.48% near the 84.51% bfloat16 (BF16) baseline; under stochastic number generator (SNG) reuse, MAC bias stays near zero while root-mean-square error (RMSE) and stochastic cross-correlation (SCC) grow as predicted.