Frontier supercomputer hits new highs in third year of exascale

Two-and-a-half years after breaking the exascale barrier, the Frontier supercomputer on the Division of Power’s Oak Ridge Nationwide Laboratory continues to set new requirements for its computing pace and efficiency.

The HPE Cray EX supercomputing system reported new highs for problem-solving speeds this week, up to date for the TOP500 announcement on the International Conference for High Performance Computing, Networking, Storage and Analysis, or SC24, in Atlanta. The Frontier crew achieved a Excessive-Efficiency Linpack, or HPL, rating of 1.35 exaflops, or 1.35 quintillion calculations per second utilizing double-precision arithmetic, the 64-bit worldwide scientific customary for computational accuracy.

“Frontier’s new numbers replicate the great contributions of OLCF’s computing and computational consultants who know easy methods to greatest optimize bleeding-edge high-performance computing programs to serve the evolving wants of our scientific person neighborhood,” mentioned Gina Tourassi, affiliate laboratory director for computing and computational sciences at ORNL.

“They by no means cease pushing boundaries, all through the lifetime of the system.”

The rating earned Frontier the No. 2 spot on the November 2024 TOP500 record, which ranks the quickest supercomputers on the planet. Frontier debuted atop the record in Could 2022 at 1.1. exaflops as the primary machine to realize exascale efficiency at greater than a quintillion calculations per second.

“Our in-house crew of consultants understands simply easy methods to get probably the most from this method by way of efficiency,” mentioned Ashley Barker, director of the Oak Ridge Management Computing Facility, which homes Frontier. “They’re the world’s most skilled exascale computing crew, and so they ran this check on Frontier whereas the machine ran different science issues for customers.”

Frontier’s improved pace marks a bounce of roughly 150 petaflops, or 150 quadrillion calculations per second—roughly equal to the efficiency of its predecessor supercomputer Summit, which was decommissioned final week.

“We may already resolve amazingly giant issues on Frontier—the most important science issues on the planet,” mentioned Al Geist, an ORNL Company Fellow and Frontier challenge director. “This 12 months, we have principally picked up the ability of one other supercomputer (like Summit).”

Frontier depends on a constellation of 1000’s of nodes, every a self-contained supercomputer of 1 CPU and 4 GPUs, linked by greater than 90 miles of cable to allow them to speak and work collectively on giant issues. Frontier debuted in 2022 with greater than 9,400 nodes, and crews have since added one other 400 nodes for a present whole of greater than 9,800. The most recent HPL benchmark ran throughout 9,500 nodes.

Geist estimates about half of Frontier’s improved rating might be credited to the brand new nodes, which had been used for utility improvement and testing by the Exascale Computing Challenge. The ECP oversaw improvement of software program purposes for Frontier and different exascale machines, such because the Aurora supercomputer at Argonne Nationwide Laboratory and the El Capitan supercomputer at Lawrence Livermore Nationwide Laboratory, and accomplished its work earlier this 12 months.

On high of the brand new nodes and exascale expertise gained by OLCF computing consultants, Frontier boasts enhancements to its math libraries developed by HPE, which constructed Frontier, and AMD, maker of the CPUs and GPUs that energy Frontier.

Apart from the up to date HPL quantity, the Frontier crew logged a brand new Excessive-Efficiency Linpack-Blended Precision, or HPL-MxP, rating of 11.4 exaflops, or 11.4 quintillion calculations per second, operating throughout all 9,800 nodes. That is a bounce of greater than an exaflop from the earlier HPL-MxP rating of 10.2 exaflops.

The detailed simulations used to mannequin such phenomena as most cancers cells, supernovas, coronavirus or the atomic construction of components require 64-bit precision, a computationally demanding customary of accuracy. Machine-learning algorithms used for synthetic intelligence usually require much less precision—typically as little as 32-, 24- or 16-bit accuracy.

“This reveals how succesful Frontier is for AI issues,” Barker mentioned.

“The outcomes we get from mixed-precision runs are equally correct, however they’re reached another way, mathematically. For some questions, double precision will stay the usual, however at these potential speeds, we count on extra researchers will begin to discover what sort of outcomes they’ll obtain through blended precision.

“They’d wish to benefit from this capability to resolve issues quicker by an element of 10, and our crew is aware of easy methods to make that occur.”

Frontier is an HPE Cray EX system with greater than 9,800 nodes, every geared up with a third-generation AMD EPYC CPU and 4 AMD Intuition MI250X GPUs. The OLCF is a DOE Workplace of Science person facility.