UW Launches Two New Computing Clusters to Expand Research Capabilities

The University of Wyoming’s Advanced Research Computing Center (ARCC) has launched two new high-performance computing clusters, SquareTop and GannettPeak.

 

These clusters, named after mountain peaks in Wyoming, expand the university’s research computing capacity and provide researchers and students with access to advanced computational resources.

 

An Aug. 25 launch event at the UW Data Center brought together staff, faculty, students and university leaders to recognize the expanded capabilities and highlight how advanced research computing can help UW researchers tackle increasingly complex questions. UW Executive Vice President Kraig Sheetz inaugurated the clusters.

 

“UW President (Shane) Reeves is committed to make UW a premier leader in artificial intelligence (AI) research and education,” Sheetz says. “These two new clusters provide the infrastructure needed to support this goal.”

 

The clusters resulted from collaborative innovation by UW’s faculty from two colleges and staff members from ARCC and UW Information Technology.

 

“These clusters will provide computing power to support the increasing portfolio of AI research, AI startups and innovations in AI workforce development,” says Vice President for Research and Economic Development Parag Chitnis. “They will help UW lead in assisting Wyoming educational institutions in making Wyoming a leader in AI-savvy workforce.”

 

The new computing resources will give UW researchers greater access to the computing power needed to analyze large datasets, run complex simulations, develop computational models and explore emerging technologies. By expanding access to high-performance computing, ARCC is helping researchers move from questions to discoveries more efficiently and enabling new opportunities for collaboration across disciplines.

 

The clusters also provide an important resource for training the next generation of researchers. Students working alongside faculty can gain hands-on experience using advanced computational tools and methods that are increasingly important in research and the workforce.

 

“The GannettPeak and SquareTop clusters are unique in their hardware architecture, which addresses the memory latency wall faced by high-performance computing and AI researchers in using graphics processing units (GPUs). Both of these clusters use ARM processors with GPUs and shared memory, rather than x86 processors with GPUs and separate memory, which leads to higher memory latency,” says Suresh Muknahallipatna, the faculty director for research computing resources. “UW faculty and graduate students are now able to use GPUs on the two new clusters without having to spend time and effort addressing the memory latency issue, which was a major impediment to GPU adoption.”

 

While the GannettPeak cluster is suited for both HPC and AI research, SquareTop is specialized for AI research, particularly large language model research and development.

 

ARCC supports UW’s research enterprise by providing computing resources, expertise and services that help researchers make the most of computational methods and data-intensive research. The center works with researchers across the university to help them access and effectively use advanced computing resources for their research.

 

The new clusters strengthen UW’s research infrastructure and position the university to pursue increasingly ambitious research questions while supporting collaboration, innovation and discovery across disciplines.

 

“While the deployment of these two new computing clusters supplies critical physical hardware, the true value of this infrastructure is ultimately about the people it empowers,” says Andrew Kirby, principal investigator of the National Science Foundation Major Research Instrumentation Program project and an assistant professor in the School of Computing and the Department of Mathematics and Statistics. “By serving as a shared foundation for innovation, these systems will catalyze essential collaborations, strengthen vital relationships and accelerate the widespread dissemination of transformative research results.”

 

As computational research continues to grow, investments in advanced research computing will play an increasingly important role in UW’s ability to generate new knowledge, develop innovative solutions and address challenges facing Wyoming and beyond.

 

About the Advanced Research Computing Center

 

UW’s Advanced Research Computing Center provides high-performance computing resources and support to help UW faculty, students and researchers conduct computationally intensive research. ARCC supports research across disciplines and helps build computational expertise and capacity throughout the university.

 

About the GannettPeak Cluster

 

The GannettPeak cluster is a powerful new supercomputer made up of 36 computer nodes and 60 GPUs. It is designed to handle large and complex research projects, including AI. Funded by a National Science Foundation Major Research Instrumentation Program grant, the system can perform up to 4.23 petaflops for highly precise scientific calculations and up to 118.74 petaflops for AI work. A petaflop means the computer can perform one quadrillion calculations per second. GannettPeak also has 5.76 terabytes of high-speed memory, allowing researchers to work with very large amounts of data and train advanced AI models.

 

 About the SquareTop Cluster

 

The SquareTop cluster is a new type of computer system made up of 24 powerful nodes, with the ability to grow to 64 nodes in the future. Each node has a specialized chip designed for AI, giving the system a total of 24 GPUs. SquareTop can provide up to 24 petaflops of computing power, making it especially useful for developing and running large language models, the technology behind many AI tools. One of the unique features of SquareTop is that the computers can work together as one large system to train AI models, or each computer can work on its own to run AI applications. UW is one of only a small number of universities to build and set up a system like this in house.