MRI: Acquisition of a High Performance Hybrid Computer Cluster for Computational Modeling

This project will permit the purchase, installation, and operation of a new high-performance computing (HPC) resource, called CRUNTCH4, to be deployed at the University of North Texas' (UNT) Center for Advanced Scientific Computing and Modeling (CASCaM). This much needed HPC resource combines different computing architectures and significant amounts of data storage, all connected via a high-speed communications fabric. This computing resource will provide the means for CASCaM investigators to continue research on a broad range of topics including quantum chemistry, materials design, biomolecular simulations, machine-learning based chemical discovery, and bioinformatics, among others. CASCaM hosts a robust computational modeling program composed of computational and experimental researchers involving development and application of computational tools in an interdisciplinary environment encompassing Chemistry, Physics, Biology, Materials Science and Engineering. UNT's CASCaM program has a strong emphasis on research and research training of early career scientists – Ph.D., undergraduate, and high school students. The CASCaM researchers comprise 40% women and 20% minoritized peoples. Additionally, UNT Chemistry has one of the longest-running NSF-REU summer scholar programs in Chemistry, with many participants using CASCaM computing resources.

The HPC resource design and capabilities will enable current and future CASCaM researchers to conduct state-of-the-art simulations, addressing a diverse array of important scientific problems. The hybrid architecture of CRUNTCH4 is composed of a 54-node HPC system coupled to a large storage device and medium-term backup storage device with high-speed interconnects. This system comprises 43 general purpose Intel Xeon CPU nodes (36 core, 196 GB mem/node), 7 Nvidia Ampere graphics processing unity (GPU) nodes (4 GPU/node with NVLink), 2 large memory CPU nodes (48 core,1.5 TB mem/node), two administration nodes, one 1 PB ExaScaler Network-attached Storage (NAS), and one 670 TB block storage array node, all connected via HDR100 InfiniBand fabric, resulting in over 484 TFLOPS of peak theoretical performance. The heterogeneous architecture of this resource will enable the investigation of a broad range of scientific topics, including high-level quantum chemistry, hybrid quantum mechanics/molecular mechanics, materials design, biomolecular simulations, bioinformatics, machine-learning based chemical discovery, to name a few. The diversity of career experience and research topics in CASCaM makes for an exhilarating scientific environment, and thus the impact of the equipment across different scientific fields and career stages will be tremendous.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.