As details about the Frontier supercomputer emerge, the US Department of Energy’s (DOE’s) Oak Ridge Leadership Computing Facility (OLCF) is seeking partnerships with select applications teams to develop scientific applications for highly effective use on the Frontier system.
Through its Center for Accelerated Application Readiness (CAAR), the OLCF will partner with simulation, data-intensive, and machine learning application teams consisting of application core developers and OLCF staff members. The teams will receive technical support from Cray and AMD—Frontier’s primary vendors—and have access to multiple early-generation hardware platforms in the run up to the system’s 2021 delivery.
Frontier, Oak Ridge National Laboratory’s first exascale system, will support the DOE Office of Science in its broad science and energy mission, advancing knowledge in critical areas of government, academia, and industry.
“Dating back to 2008 and the deployment of the Titan supercomputer, CAAR has served as a successful model for preparing accelerated applications for one-of-a-kind systems unlike any of their predecessors in terms of size and scale,” said Bronson Messer, OLCF CAAR director. “We’re looking forward to renewing this program to ensure that the capability to perform exascale science is awaiting Frontier upon its arrival.”
Based on Cray’s new Shasta architecture, Frontier will feature AMD EPYC CPU and Radeon Instinct GPU technology and sport a peak performance exceeding 1.5 exaflops. Each node on Frontier will contain one CPU and four GPUs bridged by coherent memory and high-speed AMD Infinity Fabric links between CPUs and GPUs. Internode communication will be facilitated by Cray’s Slingshot interconnect.
Leading up to the delivery of Frontier, the CAAR application teams will redesign, port and optimize their software to the system’s architecture and demonstrate the effectiveness of their applications through a scientific grand-challenge project.
CAAR partnership project proposals, accepted now through June 8, will be evaluated by a computational and scientific review conducted by the OLCF. In addition to gauging the scientific merit and acceleration plan of each proposal, the committee will strive to select a mix of computational algorithms and programming approaches representing a broad range of scientific disciplines.
For more information on applying to CAAR, visit: https://olcf.ornl.gov/frontier
