The Future of Computing beyond Moore's Law
John Shalf
Abstract
Society has come to depend on the rapid, predictable, and affordable scaling of computing performance for consumer electronics, the rise of “big data” and hyperscale data centers, scientific discovery, and national security. Quantum, neuromorphic, and artificial intelligence (AI) technologies have garnered increasing attention recently due to the rapid pace of recent improvements. Much of advanced architecture development and many new startup companies in the digital computing space are targeting the AI/ML market because of its explosive market growth rate. New software implementations, and in many cases new mathematical models and algorithmic approaches, are necessary to advance the science that can be done with new architectures. Extracting more compute performance alone may not be sufficient to realize performance gains in future systems. Architectural specialization is creating new data center requirements such as support for emerging accelerator technologies for machine-learning workloads and rack disaggregation strategies that will push the limits of current interconnected technologies.