This paper proposes an energy-efficient deep inmemory architecture for NAND flash (DIMA-F) to perform machine learning and inference algorithms on NAND flash memory. Algorithms for data analytics, inference, and decision-making require processing of large data volumes and are hence limited by data access costs. DIMA-F achieves energy savings and throughput improvement for such algorithms by reading and processing data in the analog domain at the periphery of NAND flash memory. This paper also provides behavioral models of DIMA-F that can be used for analysis and large scale system simulations in presence of circuit non-idealities and variations. DIMA-F is studied in the context of linear support vector machines and knearest neighbor for face detection and recognition, respectively. An estimated 8×-to-23× reduction in energy and 9×-to-15× improvement in throughput resulting in EDP gains up to 345× over the conventional NAND flash architecture incorporating an external digital ASIC for computation.

A planar floating-gate NAND technology has previously realized a 0.87Gb/mm2 memory density using 3b/cell [1] and achieved a minimum feature size for 16nm [2]. However, the development of planar NAND flash is expected to reach the scaling limit in a few technology generations. To break though this limit, a significant shift to 3D NAND flash has begun and several types of 3D memory cell structures have been proposed and discussed [3–5]. Recently a 3D V-NAND technology achieved 1.86Gb/mm2 using charge-trap cells and 3b/cell [6]. This paper presents a 3b/cell NAND flash memory utilizing a 3D floating gate (FG) technology that achieves 4.29Gb/mm2.

The aggressive scaling of NAND Flash memory technology - one that is even outpacing Moore's Law - has enabled very rapid cost-per-bit reduction, resulting in an explosion of systems utilizing this versatile memory technology. From removable media and personal music players to smart phones, tablets, and now personal computers and data center applications employing client and enterprise solid state drives (SSDs), NAND technology is making solid-state memory-based storage affordable.