The Frontier supercomputer at the Department of Energy’s Oak Ridge National Laboratory earned today’s top rating as the world’s fastest in the TOP500 list #59, with a performance of 1.1 exaflops. The system is the first to achieve an unprecedented level of computing performance known as the exascale, a threshold of quintillion arithmetic operations per second.
The Frontier features a theoretical peak performance of 2 exaflops, or 2 quintillion calculations per second, making it 10 times more powerful than ORNL’s Summit system. The system leverages ORNL’s extensive experience in accelerated computing and will enable scientists to develop technologies desperately needed for the country’s energy, economy, and national security, helping researchers tackle problems of national importance that were impossible to solve just five years ago.
“Frontier is ushering in a new era of Exascale computing to solve the world’s greatest scientific challenges,” said Thomas Zakaria, ORNL Director. “This achievement offers just a preview of Frontier’s unparalleled ability as a tool for scientific discovery. It is the result of more than a decade of collaboration between national laboratories, academia, and private industries, including the Department of Energy’s Exascale Computing Project, which deploys applications, software technologies, hardware, and the integration necessary to ensure The effect on the exascale.”
Ranking announced in International Supercomputing Conference 2022 in Hamburg, Germany, which brings together leaders from around the world in the field of High Performance Computing, or HPC. Frontier speeds exceed those of any other supercomputer in the world, including ORNL’s Summit, which is also located at ORNL’s Oak Ridge Command Computing Facility, a DOE Office of Science user facility.
Frontier, the HPE Cray EX supercomputer, was also ranked number one on the Green500 list, which ranks energy use and efficiency by commercially available supercomputer systems, with a performance of 62.68 gigaflops per watt. Frontier closed the rankings twice a year with first place in a newer category, Mixed-Fidelity Computing, which ranks performance in formats commonly used for artificial intelligence, with a performance of 6.88 exaflops.
Work on Frontier delivery, installation and testing has begun during the COVID-19 pandemic, as lockdowns around the world strain international supply chains. More than 100 members of the public and private sector team worked around the clock, from sourcing millions of components to ensuring system parts were delivered on schedule to carefully installing and testing 74 HPE Cray EX supercomputers, which include more than 9,400 operating nodes Powered by AMD and 90 miles of network cable.
“When researchers are able to get to the fully operational Frontier system later this year, it will be the culmination of work that began over three years ago with hundreds of talented people across the Department of Energy and our industry partners at HPE and AMD,” ORNL Associate Director Computing and Science Lab Jeff Nichols. “Scientists and engineers from around the world will put these extraordinary computing speeds to work solving some of the most challenging questions of our time, and many will begin their exploration on day one.”
Frontier’s overall performance of 1.1 exaflops translates into more than quintillion floating-point operations per second, or fluctuations, as measured by Linpack Benchmark High Performance Test. Each failure represents an arithmetic operation, such as addition, subtraction, multiplication, or division.
Frontier’s early performance on the Linpack benchmark is more than seven times that of Summit’s at 148.6 petaflops. Summit continues as a great, highly-rated giant machine for open science, listed at number four on the TOP500.
Frontier’s mixed-precision computing performance scored approximately 6.88 exaflops, or more than 6.8 quintillion flops per second, as measured by Introspection Linpack-Accelerator High Performanceor the HPL-AI test. The HPL-AI test measures computational speeds in computing formats typically used by machine learning methods that drive advances in artificial intelligence.
Detailed simulations that traditional HPC users have relied on to model phenomena such as cancer cells, supernovae, coronavirus or the atomic structure of elements require 64-bit precision, a computational form that requires computing precision. Machine learning algorithms typically require much lower resolutions—sometimes as high as 32, 24, or 16-bit—and can take advantage of the proprietary hardware of GPUs, or GPUs, that machines like Frontier rely on to reach higher speeds.
ORNL and its partners continue to implement the Frontier program on schedule. Next steps include continued testing and validation of the system, which remains on track for final acceptance and early science access later in 2022 and open to full science at the beginning of 2023.
Facts about Frontier
Exascale performance for the Frontier supercomputer is enabled by some of the world’s most advanced technologies from HPE and AMD:
- Frontier contains 74 HPE Cray EX supercomputer cabinets, which have been specifically designed to support the performance and scale of next-generation supercomputing, once opened for early access to science.
- Each node contains an enhanced EPYC processor and four AMD Instinct accelerators, for a total of more than 9,400 CPUs and more than 37,000 GPUs across the entire system. These nodes provide developers with easier capabilities to program their applications, due to the coherence provided by EPYC processors and Instinct accelerators.
- HPE Slingshot, the world’s only high-performance Ethernet fabric designed for next-generation HPC and AI solutions, including larger data-intensive workloads, to meet the demands of higher speed control and congestion for applications to run smoothly and boost performance.
- An I/O subsystem from HPE that will come online this year in support of Frontier and OLCF. The I/O subsystem features an in-system storage layer and Orion, an enhanced center-level file system based on Luster that is also the world’s largest and fastest parallel file system, based on the Cray ClusterStor E1000 storage system. The storage layer within the system will use compute node local storage devices connected via PCIe Gen4 links to provide maximum read speeds of more than 75TB/s, maximum write speeds of more than 35TB/s, and more than 15 billion randomizations. Read I/O operations per second. The Orion large-scale file system will provide approximately 700 petabytes of storage and maximum write speeds of 5 terabytes per second.
- As the world’s fastest next-generation giant computing system for open science, Frontier is also energy efficient, due to its liquid-cooling capabilities. This cooling system promotes a quieter data center by removing the need for a noisier air cooling system.
Supercomputers emerge as the world’s fastest Exascale barrier (2022, May 30)
Retrieved 30 May 2022
This document is subject to copyright. Notwithstanding any fair dealing for the purpose of private study or research, no
Part may be reproduced without written permission. The content is provided for informational purposes only.