Category I: Stampede3 -- Modernizing and Evolving The Largest ACCESS Compute Resource

Computing, for simulation or data analysis, is a fundamental enabler of all fields of engineering and science supported by the National Science Foundation. The Stampede and Stampede2 projects deployed enormously successful supercomputers that have delivered billions of processor hours of computer time to thousands of NSF funded projects, enabling key insights in diverse areas, from engineering the latest drug discoveries to studying the formation of the earliest galaxies and supermassive black holes. This proposal from the Texas Advanced Computing Center (TACC) at The University of Texas at Austin will continue the Stampede legacy and deploy a new comprehensive supercomputing environment with the latest processors and accelerators with high-bandwidth memory technologies. The Stampede 3 system nodes will be more than quadruple the performance of the current Stampede 2 nodes in many of the dimensions that are most important to scientific computing: computing capability, memory bandwidth and capacity, and network bandwidth. Stampede 3 will not only serve today’s researchers but will continue the community on an evolutionary path toward many-core processors and accelerated technologies, better preparing the community for the next generation of supercomputer systems. Stampede 3 is designed around Intel’s Xeon Max processors with high-bandwidth memory and a small number of Xeon Max GPUs, while also leveraging a portion of the most recent Stampede 2 hardware. Stampede 3 will maintain the familiar programming model for all of today’s users, and thus will be broadly useful for traditional simulation users, users performing data intensive computations, and emerging classes of new users. The project will continue the successful partnership between Dell, Intel, and TACC. TACC and its partners have unmatched expertise in operating large scale computing systems for the scientific community. In addition to supporting the system, the project will continue to offer 24x7 operations and superior support from High-Performance Computing experts familiar with a broad range of scientific applications. Through its lifetime, Stampede 3 will be a critical large scale research instrument for 10,000+ of the nation’s scientists and engineers, deliver groundbreaking innovation, and be a platform for thousands of young researchers to master computational engineering and science skills at scale. The system will be deployed in the second half of 2023 and is scheduled to operate from early 2024 until early 2029. 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.