Intelligent Architectures for Intelligent Machines - Data-Centric Computing Systems

Watch a 33-minute conference talk from ETH Professor Onur Mutlu at INSAIT 2022 exploring how modern computing architectures are bottlenecked by data movement challenges. Examine three key shortcomings of current architectures in handling data, leveraging big data, and exploiting application data semantics. Learn about designing more intelligent architectures based on data-centric, data-driven, and data-aware principles. Discover novel approaches to accelerate machine learning and enable computation closer to data, including performing parallel bulk operations in memory and utilizing 3D-stacked memory technology's logic layer. Understand the importance of adopting these intelligent architectural approaches for improving computing efficiency, performance and sustainability, concluding with guiding principles for future system designs.