REMNA: Variation-Resilient and Energy-Efficient MLC FeFET Computing-in-Memory Using NAND Flash-Like Read and Adaptive Control

Nonvolatile memory (NVM)-based computing-in-memory (CiM) has shown promising prospects in deep neural network (DNN) inference at the edge thanks to its nonvolatility and high density. Moreover, most NVMs support multi-level cell (MLC) storage, which can further boost energy efficiency and storage density. However, MLC NVM-based CiMs suffer from degraded accuracy due to device nonidealities, including large variations, nonlinear current distribution, and state drifts. Although prior works have explored various mitigation measures, such as hybrid SLC/MLC, write-and-verify, and local recovery units, the substantial costs from software support, energy, latency, and area still limit the performance. Therefore, the tradeoff between inference accuracy, storage density and compute density has become a vital challenge in NVM-based CiMs. To break the inevitable tradeoff, this work proposes REMNA, a variation-resilient and energy-efficient CiM based on MLC ferroelectric field-effect transistors (FeFETs). For the first time, a NAND flash-like read scheme is introduced in MLC CiMs, enabling REMNA with reliable computing, efficient storage, and high scalability. To further enhance resilience against variations and drifts, an adaptive bias voltage control method is explored. Furthermore, an adaptive sparsity-aware input control is proposed to improve throughput and energy efficiency in the segmented REMNA architecture. Results show that REMNA achieves 1.39-18.12× storage density and 1.34-20.90× compute density with 26.72TOPS/W energy efficiency and <1% accuracy loss under 8-bit activation and 8-bit weight precision compared with the time-domain MLC FeFET CiM, multimode RRAM CiM, refined STT-MRAM CiM, and hybrid PCM CiM.