R&D: SLC and MLC In-Memory-Approximate-Search Solutions

2023 IEEE International Memory Workshop (IMW) has published an article written by Po-Hao Tseng, Tian-Cig Bo, Yu-Hsuan Lin, Yu-Chao Lin, Jhe-Yi Liao, Feng-Ming Lee, Yu-Yu Lin, Ming-Hsiu Lee, Kuang-Yeu Hsieh, Keh-Chung Wang, and Chih-Yuan Lu, Macronix International Co., Ltd., 16 Li-Hsin Rd. Hsinchu Science Park, Hsinchu, Taiwan, ROC.

Abstract: “We proposed a novel in-memory-approximate-search (IMAS) method from single-level cell (SLC) to multi-level cell (MLC) based on our commercial 48-layer (48L) and 96-layer (96L) 3D-NAND flash technology. The method with word-line input as search word provides ultra-high parallel searching capability with the database stored in high-density 3D NAND-flash IMAS chip(s). The input search word can be compared with 128 K data words in just one read cycle, where the Hamming distance (HD) computation (for SLC approach) or similarity computation (for MLC approach) can be carried out within the memory array. Latest 96L 3D-NAND IMAS chip with CMOS under array (CuA) technology provides small chip size, long search/data word, excellent output resolution, and wide matching criteria for the tolerance on sensing current variation. Application on Memory-Augmented Neural Network (MANN) was demonstrated on VGGface2 dataset, with the comparisons between the 48L and 96L SLC 3D-NAND IMAS technologies.“