Usually, this would have been the time when we are all preparing to get on our flight to Frankfurt for the annual International Supercomputing Conference. ISC is one of the main events for the HPC industry and a great opportunity for the community to meet in person, especially here in Europe. Like many other events, this year ISC will be virtual. As we will not have the pleasure to meet in person, it is a good time to share some thoughts about the status of supercomputing in Europe.
June means Top500, the list of fastest supercomputers in the world published twice a year during ISC and SC. And with the Arm-based Fugaku installation completed few weeks ago at Riken institute in Japan, it is interesting to keep an eye on that and see how the new system will rank. Even if Fugaku is not expected to hit the Exaflop, it is a contender for the first place, which may be announced by the time you read this. Certainly, it represents the fastest Arm-based system so far and an excellent example of co-design. Fugaku only relies on the A64FX CPU to deliver its performance, a CPU specifically designed by Fujitsu for HPC applications. A quite different approach from the top US system that heavily relies on accelerators, and the planned Aurora, Frontier, and El Capitan expected to be the first exascale machines in US.
We are also starting to see first deployments of A64FX machines outside Japan, with delivery planned for the second half of the year in UK as part of the Isambard2 system and at Sandia National Laboratories in US. And more is to come.
On the European side, 2020 is the year of the pre-exascale machines with systems expected to be delivered as part of the EuroHPC Joint Undertaking at the Barcelona Supercomputing Centre (Spain), CINECA (Italy) and CSC (Finland) between the end of 2020 and beginning of 2021. These systems are expected to cover a wide range of technologies preparing the field for the European Exascale program, currently planned for 2023 and aimed to adopt technology developed with the European Processor Initiative (EPI). In this context we are pleased to report that SiPearl, the company that is designing the microprocessor for the European exascale supercomputer, announced that it licensed next-generation, Secure, and scalable Arm Neoverse “Zeus” core to build its high-performance CPU.
The ability to design specialized architectures, together with hardware and software codesign, is becoming a key strategy to overcome the limit of traditional multiprocessor architectures and reach the exascale level performance required for the next generation of HPC systems. Another trend is to leverage HPC as a component of a more complex workflow that involves data generated at smart sensors in an IoT environment with the flexibility to process this data at the edge or move it to a centralized HPC cloud center as well as anywhere in between in so called “Fog nodes”. Arm is playing a fundamental role making silicon design faster and cheaper, providing affordable access to a huge range of market leading technologies, from HPC and cloud to the edge.
On the product side, the offering has never been so rich, with a long list of Arm-based options now on the market. In addition to the Fujitsu A64FX chip, available on both Fujitsu FX700/1000 and Cray machines and the Huawei Kunpeng 920 announced last year, other exciting announcement came out earlier this year. Ampere revealed its 80 core Ampere Altra CPU with an impressive support for 128 lanes of PCIe Gen4 per socket and Marvell announced its 96 core ThunderX3 processor with similarly impressive memory bandwidth and I/O capabilities. On the cloud side, AWS Graviton2 demonstrated to be a very competitive option for HPC workloads, making a solid case for HPC in the cloud as discussed from Arm Sr. Director for HPC Brent Gorda in his blog.
Finally, the global COVID pandemic shined a spotlight on HPC demonstrating how this technology is a fundamental tool for scientific advancement and the impact it can have on our lives. Under the EU PRACE COVID-19 Initiative, HPC centers and research infrastructures across Europe opened their facilities providing priority access to research for COVID-related studies. Activity has been also coordinated with the COVID-19 High-Performance Computing Consortium, the US counterpart composed by major American industrial and research centers, along with the contribution of the UK Digital Research Infrastructure. This is an excellent example of how the industry can collaborate for a common good.
At Arm, we remain focused on HPC providing leading edge technologies to our partners and enabling them to develop the next generation of supercomputers. We also continue to invest in Europe with a dedicated team of application engineers focused on supporting final users to port and optimize their application on Arm. We are involved in collaboration with major European centres of excellence and with several universities across EU and UK. Personally, I am pleased to lead the HPC business in EMEA and look forward to meeting most of you in person soon.
Daniele leads Arm HPC business in EMEAI region, supporting partners and users in driving the adoption Arm technologies for the next generation of supercomputers. Daniele spent most of his career in IT infrastructure business, with a special focus on HPC and Cloud technologies, covering both technical and commercial roles. He has been involved in design and deployment of some of the largest HPC installations in Europe for scientific research and industrial applications.
[CTAToken URL = "mailto:daniele.piccarozzi@arm.com" target="_blank" text="Contact Daniele Piccarozzi" class ="green"]