Monolithic 3d Compute In Memory
Notre Dame, Georgia Tech, IEDM 2020 paper Highlighted in Semiconductor Engineering, Analog Memory, Neuromorphic Applications, and Other Ideas
Today, billions of connected edge devices produce zettabytes of data that need to be transformed into actionable information. This has created an unprecedented demand for data-centric computing. Compute-in-memory (CIM) is a promising approach to overcome memory bottleneck where compute is moved closer to the data residing in the memory. In last year’s IEDM 2020 conference, the team of researchers at Notre Dame led by Dr. Sourav Dutta from Prof. Suman Datta’s group demonstrated for the first time a monolithic 3D integration solution that can immensely accelerate compute-in-memory. Their work demonstrates a fully back-end-of-line (BEOL) compatible ferroelectric field-effect transistor (Fe-FET) with low temperature processing, ultra-scaled channel length, ultra-fast write speed, high endurance cycle and multi-bit programming capability. With such a monolithic 3D architecture, the researchers exhibit a significant advantage in area, energy, and latency compared to conventional 2D architecture.
IEDM paper: "Monolithic 3D Integration of High Endurance Multi-Bit Ferroelectric FET for Accelerating Compute-In-Memory” https://ieeexplore.ieee.org/abstract/document/937197
4
