DRAM scaling is pivotal for big AI computing era, with 3D stackable DRAM [1]-[3] considered a solution in the future. Positioned between conventional 2D DRAM and future 3D DRAM, the 4F 2 DRAM [4] is an attractive candidate. In this work, we propose a stacked 4F 2 DRAM with circuit under array (CuA) to enhance the memory density. To avoid new channel material complexities, we propose adopting an annealed poly-Si channel access transistor (akin to [5]), with an additional auxiliary gate (AG) to minimize the GIDL leakage current. Results show that the Si-based channel material with gate-all-around (GAA) transistor can achieve ultra-low leakage current ~0.33fA at drain bias =1V, supporting >10sec retention with a 7fF cell capacitance. No floating-body effect or Vt instability is observed for the nano-wire fully-depleted GAA device. The proposed stacked 4F 2 DRAM with CuA holds potential for equivalent <10nm node DRAM. The CuA process offers an opportunity to boost internal bandwidth by designing numerous small tiles for maximum parallelism, making it a promising advancement in the realm of DRAM scaling for the AI era.
No takes yet. Share an insight, caveat, or question.
Lue et al. (2024) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: