Memory-Centric Computing - Enabling Fundamentally Efficient and Intelligent Machines

Explore memory-centric computing architectures in this remote conference talk that addresses the fundamental data bottlenecks plaguing modern computing systems. Examine three major shortcomings of current computers in handling data, leveraging vast datasets, and exploiting semantic properties of application data. Learn about three key architectural design principles: data-centric, data-driven, and data-aware approaches that can dramatically improve computing efficiency and performance. Discover two promising research directions for reducing memory latency and energy consumption: processing using memory, which exploits operational properties of memory chips for massively-parallel in-memory computation, and processing near memory, which integrates sophisticated processing capabilities in memory chips or controllers. Understand how these architectures can deliver order-of-magnitude improvements in performance and energy efficiency for critical workloads including artificial intelligence, machine learning, graph analytics, database systems, video processing, climate modeling, and genome analysis. Gain insights into adoption strategies for these intelligent architectures and explore future research opportunities in computing architecture design focused on efficiency, performance, and sustainability.