US deploys mini supercomputer to test software for AMD’s Frontier


The US Department of Energy has presented a small computing cluster that will serve as a test platform for its upcoming Frontier supercomputer, which is set to become the country’s first exascale machine.

Oak Ridge National Laboratory (ORNL) said the researchers are using the “Crusher” system to test software they plan to run for Frontier, which is on its way to becoming one of the world’s leading supercomputers. Frontier is designed to deliver up to 1.5 trillion (1,500,000,000,000,000,000) operations per second, or 1.5 exaFLOPS (EFLOPS). This is more than three times the peak performance of the Japanese Fugaku, the current world No. 1.

Crusher is a miniature Frontier, which was on its way to becoming the world’s best supercomputer when it was scheduled to launch last year. But the final $600 million system is still being integrated and tested at ORNL.

Crusher is based on the same architectural building blocks as the Frontier supercomputer, which will consist of 100 HPE-designed server cabinets. Crusher, however, consists of just 1.5 cabinets, combining 192 server nodes connected by HPE’s Slingshot Interconnect. Each node contains a third-generation 64-core AMD EPYC processor paired with 512GB of DDR4 memory and four AMD Instinct MI250xGPUs powered by 512GB of HBM2e.

The Crusher system contains a large amount of computing power although it was designed for troubleshooting and testing software to run on Frontier in the future. ORNL said it produced more performance than its now decommissioned Titan supercomputer, the world’s fastest when it was introduced a decade ago, while taking up 1% of the floor space. Crusher takes up 44 square feet, compared to the Titan at 4,352 square feet.

Switching on the treatment

The United States is racing against China and other nations to take the lead in supercomputers. These systems are essential for research in areas ranging from the development of advanced materials, drugs and weapons to the design of automobiles and consumer goods. They can handle the heavy calculations used to model the implications of climate change and simulate galaxies, providing information on how they form and evolve.

ORNL said researchers are already seeing promising results on the Crusher test bed. He is helping them prepare the software for Frontier upon completion by the end of 2022 and availability for outside researchers next year.

The Crusher system runs a type of astrophysical hydrodynamics software used to simulate the dynamics of galaxies up to 15 times faster than the lab’s current supercomputer, the IBM-based Summit machine, which uses a combination of IBM’s Power processors. and NVIDIA Volta GPU accelerators. . Summit is the fastest system in the United States today and #2 in the Top 500 list of fastest supercomputers in the world.

The researchers also use nuclear physics software used to calculate the properties of atoms. Code runs up to 8x faster on any of the AMD Instinct MI250x GPUs powering Frontier than on Summit’s NVIDIA V100 GPUs.

Bragging rights

AMD, Intel and NVIDIA are competing for market share in high-performance server chips used in data centers. Although the production of unique supercomputer chips is only a small part of the business, the advanced technologies at the heart of these systems tend to trickle down to enterprise data centers. This gives a competitive edge to chip companies that win contracts from companies like the US Department of Energy (DoE).

AMD also landed a second $600 million contract with HPE to build an exascale supercomputer for the DoE to support the nation’s nuclear arsenal. The new supercomputer, called El Capitan, will be housed in the tightly guarded compound of Lawrence Livermore National Lab when it is completed early next year.

The United States is building another exascale supercomputer called Aurora, which promises even more performance than Frontier when it opens at Argonne National Laboratory in late 2022.

Intel, which supplies Aurora’s CPUs and GPUs, said last year that the new supercomputer would produce more than 2 EFLOPS of peak performance and deploy the same Slingshot interconnects as Crusher and Frontier.


Comments are closed.