2.5 petaflop ARM-based supercomputer commissioned by US Department of Energy
HPE Apollo 70 supercomputer will run modelling and simulation workloads for the US National Nuclear Security Administration
The world's most powerful ARM-based supercomputer to date has been commissioned by Sandia National Laboratories for the US Department of Energy.
The HPE Apollo 70 is based on the 14-nanometre Cavium ThunderX2 ARM processor. The 2.5 petaflop supercomputer, called Astra, will be deployed at the US National Nuclear Security Administration (NNSA). Part of the Vanguard project, Astra will run modelling and simulation workloads in areas such as national security, energy and science.
Mike Vildibill, vice president of HPE's Advanced Technology Group, claimed that Sandia National Labs had been "an active partner" in leveraging HPE's ARM-based platform "since its early design, and featuring it in the deployment of the world's largest ARM-based supercomputer". He added that it represented a strategic investment for the Department of Energy in the race towards exascale supercomputing.
Astra is comprised of more than 145,000 cores in 2,592 dual-processor servers with four compute nodes in a 2U form factor, with liquid cooling enabling the performance of the ARM v8 64-bit CPUs to be cranked up.
While Astra will be used by the NNSA to help manage its management and security of US nuclear weapons, nuclear non-proliferation programmes, and naval reactor programs, it is also intended to help advance ARM supercomputing.
"One of the important questions Astra will help us answer is how well does the peak performance of this architecture translate into real performance for our mission applications," said Mark Anderson, program director for NNSA's Advanced Simulation and Computing program, which funds Astra.
Scott Collis, director of Sandia's Center for Computing Research, added: "Emerging architectures come with many challenges. Since the NNSA has not previously deployed high-performance computing platforms based on ARM processors, there are gaps in the software that must be addressed before considering this technology for future platforms much larger in scale than Astra."
Sandia's Centre for Computing Research will work with its counterparts at Los Alamos and Lawrence Livermore national laboratories "to develop an improved software-and-tools environment that will enable mission codes to make increasingly effective use of Astra as well as future leadership-class platforms", said Ken Alvin, senior manager of Sandia's extreme scale computing group.
The Vanguard programme, he continued, is intended to enable the NNSA to invest in emerging technologies in a bid to broaden future supercomputing options.
In addition to working with HPE, ARM has also been "deeply engaged" with the project, as has CPU-designer Cavium.
"Vanguard is an additional proof point regarding readiness and maturity of ThunderX2 processors for large-scale deployments and will further accelerate the entire computing ecosystem on the Arm server architecture," said Gopal Hegde, vice president and general manager of the Data Center Processor Group at Cavium.
Cavium is a fab-less semiconductor company designing CPUs based on both ARM and MIPS architectures. The company agreed to an acquisition by Marvell Technology for $6 billion in cash and stock in November 2017, although the deal has yet to be completed.
The ThunderX2 line is manufactured by Global Foundries, the chip manufacturing company spun-out of AMD, on the company's 14-nanometre FinFET process.
Cavium has been able to take advantage of the cost-cutting that followed Avago Technologies' $37 billion acquisition of Broadcom (changing its name from Avago to Broadcom following the acquisition), when Broadcom ditched its Vulcan ARM server CPUs. Cavium acquired both the assets and the people responsible for them and has rolled up the technology into its ThunderX2 line.
In addition to HPE, supercomputer specialist Cray has also added an ‘ARM option' to its XC50 supercomputer range, which also supports Intel Xeon and Xeon Phi CPUs, as well as Nvidia Tesla P100 GPUs.
You may also like
/news/4333291/eu-decide-fate-usd14bn-hpe-juniper-networks-merger
Mergers
EU to decide fate of $14bn HPE-Juniper Networks merger next month
UK CMA is also investigating the deal
/news/4324071/arms-tries-block-copilot-pc-lauch
Law
Arm tries to block Copilot+ PC lauch
Part of an ongoing dispute with Qualcomm
/news/4269317/us-court-clears-mike-lynch-fraud-charges
Law
US court clears Mike Lynch of fraud charges
Lynch, who faced 15 counts of fraud, says he is ‘elated’