High-Density Multilevel 3D Vertical Resistive Switching Memory (VRRAM) for Massively Parallel in-Memory Computing

In-memory computing (IMC) can overcome the memory bottleneck of data-intensive workloads, such as deep learning on the edge. Scaling of IMC circuits is enabled by high-density crossbar arrays (CBAs) of 3D vertical resistive switching memory (VRRAM) in the back end of the line (BEOL). However, 3D-VRRAM capable of precise multilevel operation and accurate IMC has not been reported yet. This work demonstrates accurate IMC with multilevel 3D-VRRAM CBAs for both matrix-vector multiplication (MVM) and inverse/pseudoinverse matrix calculation for the first time. Energy-efficient, parallel IMC is demonstrated for real-life problems via experiments on multilevel 2D CBAs. These results support 3D-VRRAM for high-density, energy-efficient IMC in edge computing applications.