Английская Википедия:Frontier (supercomputer)

Материал из Онлайн справочника
Перейти к навигацииПерейти к поиску

Шаблон:Short description Шаблон:Use American English Шаблон:Use mdy dates Шаблон:Other uses Шаблон:Infobox custom computer

Hewlett Packard Enterprise Frontier, or OLCF-5, is the world's first exascale supercomputer. It is hosted at the Oak Ridge Leadership Computing Facility (OLCF) in Tennessee, United States and became operational in 2022. Шаблон:As of, Frontier is the world's fastest supercomputer. It is based on the Cray EX and is the successor to Summit (OLCF-4). Frontier achieved an Rmax of 1.102 exaFLOPS, which is 1.102 quintillion floating-point operations per second, using AMD CPUs and GPUs.[1][2][3][4][5]

Measured at 62.86 gigaflops/watt, Frontier topped the Green500 list for most efficient supercomputer,[5] until it was dethroned in efficiency by Flatiron Institute's Henri supercomputer in November 2022.[6]

Design

Frontier uses 9,472 AMD Epyc 7453s "Trento" 64 core 2 GHz CPUs (606,208 cores) and 37,888 Radeon Instinct MI250X GPUs (8,335,360 cores). They can perform double precision operations at the same speed as single precision.[7]

"Trento" is an optimized 3rd Gen EPYC CPU[8] ("Milan"), which is based on the Zen 3 microarchitecture.

It occupies 74 Шаблон:Convert rack cabinets.[9] Each cabinet hosts 64 blades, each consisting of 2 nodes.

Blades are interconnected by HPE Slingshot 64-port switches that provides 12.8 terabits/second of bandwidth. Groups of blades are linked in a dragonfly topology with at most three hops between any two nodes. Cabling is either optical or copper, customized to minimize cable length. Total cabling runs Шаблон:Convert. Frontier is liquid-cooled, allowing 5x the density of air-cooled architectures.[7]

Each node consists of one CPU, 4 GPUs and 4 terabytes of flash memory. Each GPU has 128 GB of RAM soldered onto it.[7]

Frontier has coherent interconnects between CPUs and GPUs, allowing GPU memory to be accessed coherently by code running on the Epyc CPUs.[10]

Frontier uses an internal 75 TB/s read / 35 TB/s write / 15 billion IOPS flash storage system, along with the 700 PB Orion site-wide Lustre filesystem.[11]

Frontier consumes 21 megawatts (MW), compared to its predecessor Summit's 13 MW.

History

The original design envisioned hundreds of thousands of GPUs and 150–500 MW of power.[7] Oak Ridge partnered with HPE Cray and AMD to build the system.

The machine was built at a cost of US$600 million. It began deployment in 2021[12] and reached full capability in 2022.[13] It clocked 1.1 exaflops Rmax in May 2022, making it the world's fastest supercomputer as measured in the June 2022 edition of the TOP500 list, replacing Fugaku.[14][15]

Upon its release, the supercomputer topped the Green500 list for most efficient supercomputer, measured at 62.68 gigaflops/watt.[5] ORNL Director Thomas Zacharia said: "Frontier is ushering in a new era of exascale computing to solve the world’s biggest scientific challenges." He added: "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."[11]

See also

References

Шаблон:Reflist

Шаблон:S-start Шаблон:S-ach Шаблон:S-bef Шаблон:S-ttl Шаблон:S-inc Шаблон:S-end Шаблон:Authority control

  1. Шаблон:Cite web
  2. Шаблон:Cite web
  3. Шаблон:Cite web
  4. Шаблон:Cite news
  5. 5,0 5,1 5,2 Ошибка цитирования Неверный тег <ref>; для сносок 2022-05-30-phoronix не указан текст
  6. Шаблон:Cite web
  7. 7,0 7,1 7,2 7,3 Шаблон:Cite web
  8. Шаблон:Cite web
  9. Шаблон:Cite web
  10. Шаблон:Cite web
  11. 11,0 11,1 Шаблон:Cite web
  12. Шаблон:Cite news
  13. Шаблон:Cite news
  14. Ошибка цитирования Неверный тег <ref>; для сносок top500 не указан текст
  15. Шаблон:Cite web