Rosie, MSOE’s supercomputer, gets major update
The rise of generative A.I. has driven a major update of Rosie, MSOE’s supercomputer, with two DGX H100s. The DGX H100 is NVIDIA’s newest supercomputing hardware—and MSOE is the first in Wisconsin to deploy it.
The DGX H100 systems deliver the scale demanded to meet the massive compute requirements of generative A.I. including large language models, recommender systems, health care research and climate science. Packing eight NVIDIA H100 GPUs per system, connected as one by NVIDIA NVLink®, each DGX H100 provides 32 petaflops of A.I. performance.
MSOE has installed two DGX H100s as well as ultra-high bandwidth InfiniBand networking to enable linking of the systems. The DGX H100s are the core computing engines behind the vast majority of headline grabbing A.I. advancement in the past 12 months. They represent the latest advancements MSOE has implemented as a leader in artificial intelligence education.
Prior to this installation, students and faculty were using Rosie to develop increasingly complex and sophisticated deep learning models for a wide variety of domains including both traditional deep learning use cases as well as the new era of generative A.I. Increasing demand for training large generative models similar to the models that drive ChatGPT has led to the need to incorporate the industry’s highest performing A.I. hardware, the DGX H100 into Rosie’s capacity.
The existing hardware in Rosie includes three NVIDIA DGX-1’s, each with eight NVIDIA Tensor Core GPUs, and 20 servers each with four NVIDIA T4 GPUs. The nodes are joined together by Mellanox networking fabric and share 300 TB of high speed network-attached storage. Rosie was designed with room for expansion to accommodate advancing technology—like the DGX H100s.
“Our undergraduate students have a unique opportunity to gain hands-on experience with Rosie,” said Dr. Derek Riley, professor and computer science program director. “Still today, there are very few universities giving their undergraduate students access to and experience with a supercomputer. At MSOE, all of our majors—from computer science and software engineering to biomedical engineering, business and nursing—are incorporating artificial intelligence technologies into the curriculum. As the first installation of DGX H100s in the state, MSOE is providing access to the most cutting-edge computing hardware available for our students and faculty. This will enable students to graduate with practical, highly demanded skills that will enable them to drive innovation in their careers for the foreseeable future.”