The U.S. Department of Energy (DOE) is preparing to develop a next-generation supercomputer that will be five times more powerful than the current fastest supercomputer, Frontier, which is expected to reach 8.5 exaflops.
The U.S. Department of Energy (DOE) aims to maintain America’s leading position in the field of high-performance computing. The DOE has issued a request for proposals for the design of an advanced supercomputer named Discover, intended to replace Frontier, the world’s fastest supercomputer as reported by Interesting Engineering on July 28. Frontier, located at the Oak Ridge National Laboratory (ORNL) in Tennessee, has topped the list of the 500 fastest supercomputers in the world for five consecutive times as of May this year.
Supercomputer Frontier at Oak Ridge National Laboratory. (Photo: ORNL).
This supercomputer is built with AMD Epyc processors, comprising 74 HPE Cray EX computer cabinets and containing nearly 8.7 million CPU cores combined with GPUs. Additionally, the machine has a Linpack performance measurement of up to 1.206 exaflops. (An exaflop is a unit of measurement for the computing power of a computer system, equivalent to one quintillion floating-point operations per second).
Although the DOE has not specified the exact performance goals for Discovery, the new supercomputer is expected to provide computing performance 3 to 5 times greater than Frontier, potentially exceeding 8.5 exaflops (a unit of measurement for the computing power of computer systems). According to Matt Sieger, the project director for Discovery at ORNL, Discovery will revolutionize scientific research across various fields, driving breakthroughs in climate change predictions, drug discovery, high-energy physics, and green energy solutions through enhanced computing power.
Discussing the potential of Discovery, Georgia Tourassi, the deputy director for computer science at ORNL, emphasized that the scientific community could model real-world scenarios at a new level of detail. “The machine will help us explore challenging issues that are not easily investigated through experimentation, observation, or theory,” Tourassi noted.
In addition to scientific applications, the new supercomputer is designed to excel in artificial intelligence and machine learning tasks, pushing the boundaries in materials science and industrial product design. Furthermore, Discovery will play a key role in the DOE’s Integrated Research Infrastructure initiative, which aims to combine various research tools and scientific facilities.
The DOE has set a deadline of August 30, 2024, for contractors to submit proposals for the Discovery supercomputer. The delivery timeframe for the Oak Ridge Leadership Computing Facility (OLCF) is set for 2027 or early 2028. Instead of specifying a precise speed target, the DOE has outlined several requirements for the next-generation system, including improved energy efficiency, full-system modeling and simulation, and enhanced AI and machine learning capabilities.
Energy efficiency remains a top priority for the OLCF, which has increased computing capacity by 500 times while only quadrupling electricity consumption over the past decade. Once operational, it is expected that researchers worldwide will have access to Discovery.
