Research Manuscript: Fast, Cool and Error Tolerant Compute-In-Memory
Event TypeResearch Manuscript
Hosted in Virtual Platform
Near-Memory and In-Memory Computing
DescriptionCompute-in-Memory (CIM) offers us a promising path to escape traditional Von-Neumann Memory Bottlenecks. This session helps push state of the art in this new paradigm by introducing ultra-low-power 3D-Nand compute substrates for AI, pioneering the deployment of ECC codes for error and variation tolerant CIM, developing high-throughput near-memory accelerators for graph workloads, and enabling ultra-fast max-min searches in DRAMs.