The world’s fastest supercomputer resides at the Department of Energy’s Oak Ridge National Laboratory (ORNL) and counts as the first true exascale machine with an HPL score of 1.102 exaflops/second.
The Frontier supercomputer was announced as the fastest supercomputer today in the 59th TOP500 list. It uses Hewlett Packard Enterprise’s (HPE) Cray EX platform, and consists of 74 purpose-built cabinets. Contained within them are a mix of AMD EPYC 64C 2GHz processors and AMD Instinct 250X professional GPUs. In total, there are more than 9,400 CPUs and 37,000 GPUs for a total core count of 8,730,112.
The huge amount of processing performance achieved equates to 52.23 gigaflops/watt and more than 1 quintillion calculations per second. That’s combined with 700 petabytes of storage and HPE Slingshot high-performance Ethernet for data transfers.
In order to cool the system, HPE pumps 6,000 gallons of water through Frontier’s cabinets every minute using four 350-horsepower pumps.
To put this performance leap in context, the previous fastest supercomputer is the Fugaku system installed at the RIKEN Center for Computational Science (R-CCS) in Kobe, Japan. It contains 7,630,848 cores and has a HPL benchmark of just 442 petaflops/second compared to Frontier’s 1.1 exaflops/second. Fugaku also offers nearly three times the processing power of the supercomputer in third place.
Frontier has a theoretical peak performance of 2 exaflops, and ORNL Director Thomas Zacharia says it will be put to very good use:
“Frontier is ushering in a new era of exascale computing to solve the world’s biggest scientific challenges. This milestone offers just a preview of Frontier’s unmatched capability as a tool for scientific discovery. It 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.”
ORNL is currently carrying out testing and validation of Frontier, with early science access to the system expected later this year. Full science access will start early next year.
