Arm technology is now powering the world’s fastest supercomputer
You may not know that Arm’s journey in HPC began nearly a decade ago. From its beginnings in the early 2010’s with the Mont-Blanc project in which the Barcelona Supercomputing Center first studied the potential of Arm-based clusters for HPC to today’s announcement at ISC2020 that the Fugaku supercomputer has achieved the top spot of the Top500 List, we’ve come a long way. In just the last eighteen months, we’ve seen significant momentum across our ecosystem:
- Arm’s first system on the Top 500 list, the Astra from Sandia National Labs, landed at #150 and was announced at SuperComputing 2018
- NVIDIA has announced full CUDA 11 support for Arm which enables a new path to designing and deploying highly energy-efficient, AI-enabled exascale compute
- Fugaku has been made available to scientists around the globe for research and response to the coronavirus, as has the Arm-based Isambard2 supercomputer
- SiPearl discussed their plans to build upon our next generation Neoverse platform “Zeus”
- The Japan Aerospace Exploration Agency will leverage the Fujitsu A64FX CPU for research and data analysis
- Sandia National Labs and Cray have announced plans to deploy systems based on the Arm-based A64FX CPU from Fujitsu
- In May, AWS announced general availability of EC2 M6g instances and most recently, C6g, and R6g instances powered by Graviton2-based processors which deliver more performance, compute, and cache for memory-intensive workloads across HPC
And, we aren’t taking our foot off the gas. HPC is all about pushing the boundaries of compute, and in the era of AI and 5G we’re seeing a new wave of evolution. Compute efficiency is more important than ever, developers want more flexibility and efficiency, the cloud is driving massive change across the infrastructure, and many vendors simply just want more choice. We developed Arm Neoverse with these challenges in mind and set out to deliver the performance, efficiency, and scalability required to make Arm a real possibility for HPC, and with all the momentum we’ve seen, it’s clear our approach is gaining traction.
Across the globe we’re seeing greater investment in Arm for HPC. From the EU to Japan, there’s demand for more customization and more control in HPC capabilities. SiPearl, backed by the European Processor Initiative, has announced its intent to develop exascale-class systems based on the next generation Arm Neoverse platform and we expect additional exascale projects based on Arm due to its openness, efficiency, and software ecosystem maturity. Arguably the most important piece in this puzzle is centered on software. Our broad investments, internally and across the HPC ecosystem, seamlessly enable migrations across instructions sets, cross-platform development, profiling, and debug. Most recently we’ve added porting guides to the Arm Compiler for Linux and to Arm Allinea Studio which port and optimize applications for Arm Scalable Vector Extensions (SVE) to help accelerate applications on the Fujitsu A64FX and future Arm CPUs supporting this exciting new technology.
As you can see, we are very proud of the progress and the fact that Arm technology truly is changing the world, especially through the work we’re seeing with the Fugaku supercomputer. Located at the RIKEN Center for Computational Science in Kobe, Fugaku is the flagship supercomputer designed to carry out a number of applications that address both social and scientific issues with the aim to help Japan achieve “Society 5.0”. The research it will enable spans drug discovery, weather and climate forecasting, new design and production processes, and in the battle against COVID-19, to gain a better understanding of diagnostics and therapeutics of the novel coronavirus.
There’s certainly more work to be done, but with Fugaku having been crowned the world‘s most efficient supercomputer in 2019 and today’s No. 1 achievement on the Top500 list, we’re seeing incredible proof of the innovation that is possible with Arm Neoverse. Seeing an Arm-based supercomputer of this scale come to life and take the crown as the top performing supercomputer in the world is an achievement our entire server ecosystem should celebrate. When you look at the upcoming platforms on our Arm Neoverse roadmap, it is clear the best is yet to come.
We extend our congratulations to the RIKEN and Fujitsu teams and thank them for many years of partnership and collaboration.
Article sourced from: https://www.arm.com/company/news/2020/06/powering-the-fastest-supercomputer
Published on: 22.06.2020