Computational Associative Memory Powered by Ferroelectric Memory.

Discovery of ferroelectricity in thin doped Hf02 has revived the interest in ferroelectric memories because its excellent CMOS-compatibility, great scalability, and superior energy efficiency. Various types of ferroelectric memories are under consideration, including capacitor based ferroelectric random access memory (FeRAM) and transistor based ferroelectric field effect transistor (FeFET). Significant progress has been made by integrating them at advanced technology nodes. These exciting developments have made ferroelectric memories prime candidates as embedded nonvolatile memory and enable their application in compute-in-memory (CiM) accelerators. One important class of CiM kernel is content addressable memory (CAM), where the memory is addressed through its content. In this work, we overview different designs of FeFET based CAM and propose FeRAM based CAM.