Researchers Document the Evolution of Commercial AI Hardware
A team from the Lincoln Laboratory Supercomputing Center has been conducting comprehensive surveys examining the performance and power changes of commercial AI accelerators since 2018.
While the transformative impact of artificial intelligence technologies on industries and national security continues, developments in hardware capabilities are also being meticulously tracked. Researchers at the Lincoln Laboratory Supercomputing Center are documenting the performance and power of commercial AI accelerators over the years.
The Evolution of AI Accelerators
Specialized systems designed to accelerate artificial intelligence technologies have been undergoing an intense development process for about a decade. These hardware components enable neural networks and deep learning processes to run more efficiently.
The Lincoln AI Computing Survey Process
The Lincoln Laboratory Supercomputing Center team has been carrying out the Lincoln AI Computing Survey studies since 2018. These conducted researches have been shared with the public through six different papers to date.
Research Team and Scope Expansion
Led by Albert Reuther, the team includes Michael Jones, Peter Michaleas, Jeremy Kepner, and Vijay Gadepally. While the first paper of the series examined 57 accelerators, the latest study reached over 120 devices.
Core Metrics and Goals
Researchers base their comparisons on peak performance and power consumption metrics. Devices are classified according to chip, board, or system categories and comprehensively analyzed.