3D Ferroelectric NAND for Ultra-High Efficiency Analog Computing-in-Memory by SK hynix

3D FeNAND with Ultra-High Computing-in-Memory Efficiency: AI models containing up to trillions of parameters require substantial memory resources to handle the vast amounts of data. Energy-efficient analog computing-in-memory (CIM) devices such as 3D vertical NAND architectures are emerging as potential solutions because they offer high areal density and are non-volatile. SK hynix researchers will detail how they achieved analog computation in ultra-high-density 3D vertical ferroelectric NAND (FeNAND) devices for the first time. They used gate stack engineering techniques to improve the analog switching properties of 3D FeNAND cells, and achieved an unprecedented ≥256-conductance-weight levels/cell. The 3D FeNAND arrays improved analog CIM density by 4,000x versus 2D arrays, and demonstrated stable multiply-accumulate (MAC) operations with high accuracy (87.8%) and 1,000x higher computing efficiency (TOPS/mm²) versus 2D arrays. This work provides an efficient method to implement the processing of hyperscale AI models in analog CIM chips for edge computing applications, where speed and low power operation are the critical requirements, not extreme accuracy.

 

Above:

(1)   is a comparison of 2D and 3D arrays for analog-CIM applications.

(2)   is a TEM analysis  of the 3D FeNAND, showing (a) a top-down view of the device; (b) a cross-sectional view at low magnification; (c) a cross-sectional view at high magnification; and (d) a schematic illustration of the FeFET cells in the 3D FeNAND array.

Source: 

IEDM 2024 Paper #38.1, “Analog Computation in Ultra-High Density 3D FeNAND for TB-Level Hyperscale AI Models,” J.-G. Lee and W.-T. Koo et all, SK hynix https://www.ieee-iedm.org/press-kit