Supercomputers are highly specialized computing systems that can process large amounts of data and perform complex calculations at speeds unmatched by standard computers. They are built by linking thousands of CPU and GPU chips together and housed in special racks, with in-built cooling infrastructure. 

Most supercomputers are housed in national laboratories of major countries and built by Hewlett-Packard Enterprise (HPE), in collaboration with chipmakers. Airbus’ previous supercomputers have been built by HPE, but when they were looking to upgrade their infrastructure on this occasion, French HPC provider Bull entered the procurement process and delivered a supercomputer at a better price. 

Airbus’ new supercomputer

Although Airbus has not publicly shared any details of its new supercomputer, system provider Bull has revealed what it can under its confidentiality agreement with the aircraft manufacturer. As per Bull, the system is based on a modular design, with the supercomputer pre-assembled in containers before being shipped to Airbus’s sites. 

Based on the BullSequana XH300 rack, the Airbus supercomputer features a mix of AMD and NVIDIA CPUs and GPUs. The storage comes from IBM’s Spectrum Scale storage, whereas NVIDIA’s InfiniBand NDR (Next Data Rate) supporting 400 Gbps per port provides the interconnection. How much compute will this deliver? We do not have an official number on this, but it is definitely three times what Airbus’s previous supercomputer could manage.

As per the agreement, Bull is providing Airbus the supercomputer on an ‘HPC-as-a-Service’ model where the aircraft manufacturer will shell out only €100 million (US$116 million) over a five-year period for the entire deal. 

How will the supercomputer be used? 

While most supercomputers are located at a single site, Bull supplied Airbus with a supercomputer at two locations in two different countries. In December last year, Airbus’ Toulouse facility in France received the hardware, while another was delivered to the Hamburg facility in April this year. 

According to Bruno Lecointe, the head of HPC, AI, and Quantum Computing at Bull, the two units are connected to function as a single supercomputer. The workloads are not split across sites; they are scheduled to run on either unit. This decision is made by a batch scheduler that considers the resources available at each unit before dividing the workload. 

“This long-term strategic and technological collaboration highlights the critical role of HPC in driving innovation and breakthrough programs across the aerospace and manufacturing industries,” said Martin Matzke, head of Central Europe and Northern Europe, at Bull in a press release. 

As Airbus enters a new era of aerospace, it is relying on the supercomputer not just to help design the airframes of its helicopters and other aircraft, but also to run simulations. The supercomputer will help deliver “digital twins” of the designs that can be tested extensively in virtual environments. 

A similar application for the supercomputer will be running the CODA computational fluid dynamics (CFD) software that has been developed by Airbus, alongside the French Aerospace Lab (ONERA) and German Aerospace Center (DLR). 

In addition to delivering HPC, Bull’s Direct Liquid Cooling will also help Airbus use the heat generated from the supercomputer to keep on-site buildings warm, the press release added.