Stony Brook University’s AI Institute has acquired a state-of-the-art HGX H100 machine to train advanced machine learning models.

Stony Brook, NY, Mar 26, 2025 — The AI Innovation Institute has expanded its computing resources with the Nvidia HGX H100, a new high-performance computing system that will enhance AI3’s ability to tackle some of the most computationally demanding machine learning problems, from language, vision, and healthcare to education, sustainability, and weather modeling.

Equipped with eight NVIDIA’s HGX H100 GPUs and an impressive 6TB of memory, this powerful setup will ensure that the machine can handle large volumes of data and complex calculations at speeds that were previously difficult to achieve, without sacrificing performance.

Steven Skiena, Interim Director of AI3, said, “This is the most powerful machine in our AI3 computing cluster, and it is being kept quite busy by our students.”

A Ph.D. student, Panagiotis Kaliosis,  Department of Computer Science, said, “I found the new AI3 Cluster really useful. We are currently working with cutting-edge Large Language Models (LLMs), such as Llama and DeepSeek, which require significant computational resources that are now more easily accessible. We are now on track to submit a new paper that makes a timely contribution in the intersection of Computer Science, Psychology, and Cognitive Science.”

The machine is powered by dual AMD EPYC 9004 Series processors, offering up to 128 cores, so the system can handle multiple complex processing tasks at the same time, making it ideal for both AI and high-performance computing applications. In addition to its powerful memory, the system features 18 drive bays for quick and efficient data storage and retrieval, allowing large datasets to be accessed quickly and smoothly running operations even when working with huge volumes of information.

Peter Zeng, a Ph.D. student in the Computer Science Department, said, “The new HGX cluster has been super useful in running experiments at a speed that hasn’t typically been possible with the other computer servers at Stony Brook. I’m grateful to be able to access this level of computing for my research exploring building trust, ensuring transparency, and avoiding biases in decisions driven by AI.”

The new H100 GPU has been shown to outperform previous-generation NVIDIA GPUs by a wide margin, and is designed to work seamlessly with NVIDIA’s NVLink interconnect technology, which allows for high-bandwidth communication between GPUs, enabling users to scale up their computing performance quickly and easily, making it an ideal solution for large-scale machine learning and deep learning workloads.

This new system is a game-changer for a range of applications. It will significantly accelerate model training, supporting faster iterations and more sophisticated models. With the addition of this cutting-edge HGX H100 8-GPU machine, the AI Innovation Institute is poised to continue pushing the boundaries of AI research and technology, offering definitive computational resources to researchers looking to drive innovation and solve complex challenges in artificial intelligence and machine learning.

The upgrade to the HGX H100 8-GPU machine was partially funded by the National Science Foundation (NSF Major Research Instrumentation award by 191952), matching funds from the Institute for Advanced Computational Science (IACS) and The Office of Research Compliance (ORC) at Stony Brook, without whose vision, guidance, and support, this would not have been possible.

Ankita Nagpal
Communications Assistant