The new Frontier supercomputer will be the fastest in the world

The Frontier system, which is jointly operated by Hewlett Packard Enterprise and Oak Ridge National Laboratory, is the world's most powerful supercomputer.

According to a recent ranking by TOP500, a project that tracks and ranks 500 of the world's most powerful computers, a US supercomputer has officially gained the top place in the world.

Frontier, a Tennessee-based system managed by Hewlett Packard Enterprise, AMD, and the US Department of Energy's Oak Ridge National Laboratory, has surpassed the Fugaku system at the RIKEN Center for Computational Science (R-CCS) in Kobe, Japan, to take first place.

In a press release, Oak Ridge National Laboratory stated that this Frontier system is “the result of more than a decade of collaboration among the national laboratories, academia and private industry, including DOE’s Exascale Computing Project, which is deploying the applications, software technologies, hardware and integration necessary to ensure impact at the exascale.”

What can a supercomputer this large and complicated accomplish? According to a March presentation, some of the early Frontier research projects involve modelling Earth systems to evaluate biogeochemical cycles and power grid planning. Experts predict that when these supercomputers grow in size, they will specialize on a small number of applications.
Here's a breakdown of what Frontier can achieve in terms of statistics.

1.1 exaflops

According to Oak Ridge National Laboratory, the Frontier supercomputer has 1.1 exaflops of capability, which means it can process nearly a quintillion computations per second. That's 1,000,000,000,000,000,000, or 1018. Frontier can theoretically chart 2 exaflops (two quintillion computations per second) at peak performance. This is ten times more powerful than the previous IBM Summit system at Oak Ridge National Laboratory, which had a peak capability of 200 petaflops (200,000 trillion computations per second). Summit was skilled in analyzing and comparing genomes, as well as simulating global weather patterns with deep learning. (Learn more about the terms "flop" and "teraflop" here.)

The first true exaflop machine

According to TOP500, the Fugaku system in Japan, which had previously held the top spot on the list for two years, now has a performance of 442 petaflops per second. Despite the fact that Fugaku has the potential to break the 1 exaflop barrier at peak performance, Frontier has been proclaimed the first exascale computer. Aurora, a system developed by Intel and Argonne Laboratory, was also supposed to reach this milestone, but it was apparently delayed owing to production and technological concerns. Another key caveat: according to The New York Times, two systems in China have previously passed this milestone quietly, and their operators did not submit findings for examination to the experts who supervise TOP500.

52.23 gigaflops performance per watt

To function, supercomputers require a large amount of electricity. Frontier and the Frontier Test & Development System (TDS) at Oak Ridge National Laboratory, on the other hand, are two of the most energy-efficient supercomputers in the TOP500 group, with a power efficiency of 52.23 gigaflops per watt for the Frontier system. With a rating of 62.68 gigaflops per watt, the Frontier TDS system—a smaller version of the supercomputer used for early tests—is the most power efficient system, with just 120,832 cores and a capability of 19.20 petaflops per second.

8,730,112 cores

Frontier has a total of 8,730,112 cores capable of executing concurrent operations. Frontier is made up of HPE Cray EX supercomputers with AMD CPUs. All of the processing capacity may be put to use solving complex problems and computations, including those required by AI systems.“In addition to modeling and simulating complex scientific research, across biological, physical and chemical sciences, with higher resolution, Frontier will also enable dramatic breakthroughs in AI,” Hewlett Packard claimed in a press statement.“At an exascale speed, Frontier’s users can develop AI models that are 4.5X faster and 8X larger.”On Frontier, you can find all of the programming languages, libraries, and tools you need.

700 petabytes of storage

Frontier's storage component, named Orion, can carry around 700 petabytes of data and has read rates of up to 75 terabytes per second, write speeds of up to 35 terabytes per second, and 15 billion random-read input/output operations per second. To put it in perspective, the data collected at Oak Ridge National Laboratory is 35 times that of the Library of Congress.

74 cabinets

The vast hardware of supercomputers must be divided into distinct compartments. Frontier consists of 74 supercomputer cabinets, 9,400 AMD-based nodes, and 90 kilometers of networking cable. Each cabinet is around 8,000 pounds, or the weight of a pick-up truck. The system has roughly 9,408 CPUs and 37,632 GPUs in total (GPUs are particularly useful for running artificial intelligence software). It uses 29 megawatts of electricity. Researchers should get access to the "fully operational" Frontier before the end of the year.
Previous Post Next Post