MIT Lincoln Laboratory unveils TX-GAIN AI supercomputer for research

MIT Lincoln Laboratory has brought online TX-GAIN, a new AI-optimized supercomputer at the Lincoln Laboratory Supercomputing Center (LLSC), now ranked on the TOP500.

• Main announcement: TX-GAIN is an LLSC supercomputer optimized for generative AI and high-performance research, powered by more than 600 NVIDIA GPU accelerators, with a reported peak performance of two AI exaflops, and it came online this summer; it supports generative AI, physical simulation, and data analysis for Lincoln Laboratory and MIT campus researchers and federal projects (examples: FAA collision-avoidance simulation, DoD autonomous navigation).
• Background & implementation details: The LLSC emphasizes interactive supercomputing and houses systems in an energy-efficient data center in Holyoke, Massachusetts; LLSC teams are researching the energy needs of AI and have developed software that can reduce AI training energy by as much as 80%. Collaborations include Haystack Observatory, Center for Quantum Engineering, Beaver Works, and the Department of Air Force–MIT AI Accelerator; Rafael Jaimes (Counter–Weapons of Mass Destruction Systems Group) reported TX-GAIN enables modeling far more and larger protein interactions for biological defense efforts.