An AI supercomputer is coming to campus. The Texas A&M University System has purchased multiple NVIDIA DGX SuperPODs for $45 million. This initial investment will triple the university’s supercomputing capacity, power increasingly complex research, and provide rapid processing speeds.
“The acquisition of this GPU cluster will massively change the computational infrastructure at Texas A&M University,” said Dr. Scott Schaefer, department head of the Department of Computer Science and Engineering. “With the increase in computational power, experiments that previously took weeks will be completed within hours.”
This cluster will contain 760 NVIDIA H200 GPUs—graphics processing units. Before AI, GPUs were used for video games because of their ability to process many data streams simultaneously. This parallel processing makes GPUs ideal for depicting the high-quality images and 3D spaces in gaming, and it also allows them to efficiently process the massive amount of data needed for machine learning, a type of AI that uses algorithms to analyze data sets, make predictions about them and generate new data. ChatGPT is a prominent example of this type of generative AI.
“New machine learning approaches have made significant impacts on a variety of fields and have created new approaches to solve ever more complex problems,” said Schaefer. “This GPU cluster enables our researchers to tackle those problems and will help us attract the best faculty to Texas A&M.”
With the increase in computational power, experiments that previously took weeks will be completed within hours.
Schaefer expects that this supercomputing system will have a significant impact on research projects in many areas of computer science and engineering including machine learning, robotics, computer vision, natural language processing, generative AI, materials discovery, physics and flow simulation, and scientific computation.
“This is just the beginning of the investment needed in computation. As the demand for future computation continues to increase, further investment will be necessary,” Schaefer said.
Dr. Shuiwang Ji, professor and Truchard family chair of computer science and engineering, is also optimistic about this acquisition.
“AI and machine learning are propelling the fifth paradigm of scientific discovery, yet progress has been significantly constrained by insufficient GPU computing,” said Ji. “This substantial investment will undoubtedly enhance our ability to conduct foundational AI research and apply it to address complex challenges in science and engineering, spanning from quantum mechanics to turbulence.”
The supercomputing system will be housed at the West Campus Data Center, which will be upgraded to meet the new system’s needs.