TX-GAIA (Green AI Accelerator), the new 4.7-petaflops system built by HPE and installed at MIT’s Lincoln Laboratory’s Supercomputing Center (LLSC) in Holyoke, Massachusetts, will enable new research in machine learning, engineering and science when it enters production this fall. The system was ranked as the 51st most powerful back in June, after completing a 3.9 petaflops Linpack run, and has since added another 800 teraflops to its benchmark tally.
The new AI-focused supercomputer spans 448 Apollo 2000 nodes, equipped with 896 2nd Generation Intel Xeon Scalable Gold 6248 (20-core) processors, 896 Nvidia V100 GPUs and 172,000 GB of memory, connected by Intel’s Omni-Path network.
TX-GAIA’s size and scope make it the most powerful supercomputer in New England and the third most powerful machine used by a U.S. university. Owing to its 100 mixed-precision (peak) AI petaflops, it is the number-one university AI supercomputer, according to its backers. They say TX-GAIA will perform complex deep neural networks (DNN) and other machine learning training, as well as support more traditional modeling and simulation workloads. Targeted applications include weather forecasting, medical data analysis, autonomous systems, synthetic DNA design, and new materials and devices.
“We are thrilled by the opportunity to enable researchers across Lincoln and MIT to achieve incredible scientific and engineering breakthroughs,” said Jeremy Kepner, a Lincoln Laboratory Fellow who founded heads the LLSC. “TX-GAIA will play a large role in supporting AI, physical simulation, and data analysis across all Laboratory missions.”
TX-GAIA is installed inside LLSC’s modular data center, called an EcoPOD, part of a hydroelectrically powered site in Holyoke, Mass. LLSC is also home to TX-E1, which supports collaborations with MIT campus and other institutions, and TX-Green, a 1-petaflops Dell EMC Knights Landing Phi-based machine acquired in 2016.
Established in 1951, the federally-funded Lincoln Laboratory exists to help shape the United States’ security, defense, and technology. The Lincoln Laboratory Supercomputing Center (LLSC), founded in 2016, hosts large computing clusters to enable laboratory researchers to process large sets of sensor data, create high-fidelity simulations, and develop entirely new algorithms.
TX-GAIA will serve the needs of the Department of Defense, as its most powerful computing resource. It will also support the recently announced MIT–Air Force AI Accelerator, which draws on the expertise and resources of MIT, including the LLSC, and the Air Force to carry out fundamental research for the purposes of rapid prototyping, scaling, and application of AI algorithms and systems.