Oak Ridge National Laboratory
Oak Ridge National Laboratory
19/11/2024 | News release | Distributed by Public on 19/11/2024 15:42
Oak Ridge National Laboratory receives honors in 2024 HPCwire Editors’ Choice award
The ORBIT research team accept their award from HPCWire's Tom Tabor at SC24 in Atlanta, GA. Credit: ORNL, U.S. Dept. of Energy
November 19, 2024
Topic:
The Department of Energy's Oak Ridge National Laboratory has been recognized in the 21st edition of the HPCwire Readers' and Editors' Choice Awards, presented at the 2024 International Conference for High Performance Computing, Networking, Storage and Analysis (SC24), in Atlanta, Georgia.
ORNL won the trade publication's "Top Supercomputing Achievement" award for 2024 for the Oak Ridge Base AI Foundation Model for Earth System Predictability, or ORBIT, a groundbreaking AI spatiotemporal transformer model with 113 billion parameters that could enable fast, cheap and highly accurate weather forecasts.
Powered by 49,152 AMD GPUs on the Frontier supercomputer, ORBIT achieved a sustained computing performance of 1.6 exaflops.
The work also earned the team a finalist nomination for the Association of Computing Machinery Gordon Bell Prize for Climate Modeling, which honors innovations in applying high-performance computing to climate modeling applications.
"We are honored to receive this recognition," said ORNL's director of AI Programs Prasanna Balaprakash. "ORBIT represents a significant milestone, breaking the exaflop barrier in training large-scale spatiotemporal AI foundation models for weather prediction - an effort considerably more complex and challenging than training large language models. This achievement is a testament to the dedication of our team at ORNL and our commitment to leveraging high-performance AI to address society's most complex challenges."
Powered by 49,152 AMD GPUs on the Frontier supercomputer, the Oak Ridge Base AI Foundation Model for Earth System Predictability achieved a sustained computing performance of 1.6 exaflops. Credit: ORNL, U.S. Dept. of Energy
The coveted annual 2024 HPCwire Readers' and Editors' Choice Awards are determined through a nomination and voting process with the global HPCwire community, as well as selections from the HPCwire editors. The awards are an annual feature of the publication and constitute prestigious recognition from the HPC community. They are revealed each year to kick off the annual supercomputing conference, which showcases high performance computing, networking, storage and data analysis.
"This year, as we celebrate HPCwire's 35th anniversary covering HPC and SC, and as we are at the advent of an HPC-led AI transformation, the 2024 Readers' and Editors' Choice Awards certainly reflect these exciting events," said Tom Tabor, CEO of Tabor Communications, publishers of HPCwire. "Throughout the world, we see grand challenge problems that can only be explored and solved because of HPC, now aided by AI. Rarely do these accomplishments come to light, much less are they recognized for their contribution to society. Between our worldwide readership of HPC experts and the most renowned panel of editors in the industry, the Readers' and Editors' Choice Awards represent resounding recognition of the many deep and varied HPC accomplishments throughout the world. Our sincerest gratitude and hearty congratulations go out to all of the winners."
More information on these awards can be found at the HPCwire website (www.HPCwire.com) or on X through the following hashtag: #HPCwireRCA24.
UT-Battelle manages ORNL for DOE's Office of Science, the single largest supporter of basic research in the physical sciences in the United States. DOE's Office of Science is working to address some of the most pressing challenges of our time. For more information, visit https://energy.gov/science.