New Frontier in supercomputing

US Energy Dept.’s Frontier upsets Riken–Fujitsu’s Fugaku as the fastest supercomputer; Fugaku claims pair of other titles.

The supercomputer Fugaku, jointly developed by Riken and Fujitsu, has successfully retained the top spot for five consecutive terms in multiple major high-performance computer rankings including HPCG and Graph500 BFS (breadth-first search), and has also taken second place for the TOP500 and HPL-AI rankings. The HPCG is a performance ranking for computing methods often used for real-world applications, and the Graph500 ranks systems based on graph analytic performance, an important element in data-intensive workloads.

The results of the rankings were announced on May 30 at the ISC2022 High-Performance Computing Conference, which was held in Hamburg, Germany.

But The US Department of Energy’s Frontier supercomputer has also been crowned world’s fastest at the International Supercomputing Conference 2022 in Hamburg, Germany, and the first system to enter the exascale era of computing.

The Fujitsu-Riken supercomputer. 

The results this time were made with Fugaku’s full complement of 158,976 nodes fit into 432 racks. On HPCG, it scored 16.00 petaflops. On the TOP500, it achieved a LINPACK score of 442.01 petaflops, and on HPL-AI it gained a score of 2.004 exaflops.

Frontier—The world’s first exascale supercomputer.

The US Department of Energy partnered with Cray Inc in 2019 to build a new supercomputer at the Oak Ridge National Laboratory (ORNL) in Tennessee, which already housed the Titan and Summit supercomputers—the latter being the most powerful until it was surpassed by Japan’s Fugaku in 2020.

Where other members at the higher end of the Top500 list of supercomputers are impressively fast calculators, with Fugaku managing over 415,000 trillion floating point operations per second, ORNL’s system completely overshadows previous computational efforts by quite a margin.

The Frontier supercomputer breaks through the exaflop threshold to perform 1.102 quintillion calculations per second, measured by High Performance LINPACK benchmark tests. And mixed-precision performance clocked by the High Performance LINPACK–Accelerator Introspection test—which “measures calculation speeds in the computing formats typically used by the machine-learning methods that drive advances in artificial intelligence”—shows the supercomputer operating at around 6.88 exaflops.

The supercomputer comprises 74 purpose-built HPE Cray EX supercomputer cabinets with more than 9,400 AMD CPUs and over 37,000 GPUs, 700 petabytes of data storage offering peak write speeds of 5TB/s, energy-efficient liquid cooling that sees some 6,000 gallons (22,700+ L) of water pumped through the system every minute, and 90 miles (~145 km) of network cables. It’s reckoned to have a theoretical peak performance of 2 exaflops.

The bi-annual Green500 rankings also have a new energy-efficiency champion, with the Frontier Test & Development System—which is essentially just a single rack identical to those making up the more powerful Frontier system—taking the top spot for mixed-precision computing at 62.68 GFLOPS/W. Meanwhile, the main Frontier system sits in second place with a power efficiency of 55.23 GFLOPS/W.