SK hynix Unveils 3D FeNAND for Analog Compute-in-Memory

SK hynix has announced research results showing 3D FeNAND enabling analog compute-in-memory (A-CiM), with the work published in an IEEE paper.

A-CiM performs multiply-accumulate (MAC) operations directly in memory using analog signals, reducing data movement between storage and compute units to improve processing speed and energy efficiency. 3D FeNAND stores data via the polarization states of ferroelectric materials and controls current accordingly. Unlike conventional 3D NAND, which uses a charge-trap nitride (CTN) structure, ferroelectric materials retain their internal polarization even after the external electric field is removed.

SK hynix said FeNAND can generate multiple current states within a single memory cell, making it suitable for weight data in AI computing.