Hewlett Packard Enterprise Builds AI Supercomputer for CRIANN to Accelerate Scientific Research and Innovation

France’s HPC and AI data center will provide regional access to the new AI supercomputer “Austral” to accelerate the nation’s AI initiative and advance areas in climate modeling, biotechnology, healthcare, and materials science

The new AI supercomputer is part of ongoing efforts to advance France’s national AI strategy and was co-funded by the Normandy Region, the French State, and the European Union through the European Regional Development Fund (ERDF). Austral will help train robust AI models and run simulations and data analysis to accelerate scientific discovery and innovation in climate modeling, biotechnology, healthcare, and materials science, among other areas.

Austral builds on HPE’s ongoing support for France’s AI initiatives

Austral is the latest AI development that HPE is fueling for France. The system builds on other innovations that HPE has delivered, including the Jean Zay supercomputer for GENCI (Grand Équipement National de Calcul Intensif), the Adastra supercomputer for GENCI-CINES (Centre informatique national de l'enseignement supérieur), and “Champollion,” an AI supercomputer that HPE is offering through its center of excellence in Grenoble, France, for broader use.

“France has demonstrated its national commitment to advancing AI, which is fundamental to propel innovation for the greater good and the overall prosperity of a nation,” said Justin Hotard, executive vice president and general manager, HPC, AI & Labs at HPE. “We are proud to continue partnering with France’s national and regional centers to accelerate its AI missions by building powerful solutions using our supercomputers and AI software. Through our latest collaboration with CRIANN to deliver Austral, users across public and private organizations will gain access to tools to create reliable and explainable AI models and tackle complex simulations to make scientific discoveries and advance R&D.”

CRIANN serves as a regional resource hub for researchers and users to learn how to train large AI models on various nationally funded computing resources, such as the GENCI-CINES Adastra supercomputer and other resources supplied by EuroHPC.

“CRIANN is committed to providing high-performance computing and data services to the research and education community in Normandy, as well as to the private sector,” said Hervé Prigent, managing director at CRIANN. “With Austral, we have a powerful and flexible platform that will enable us to run a wide range of applications and workflows, from simulations and modeling to AI and data analytics. We are excited to work with HPE to build Austral and enhance our HPC capabilities.”

France gains new AI supercomputer to accelerate outcomes with best-of-breed technologies

Austral is built using HPE Cray XD2000 supercomputers and features the HPE Machine Learning Development Environment (MLDE), a machine learning development software platform to rapidly train large-scale models.

The new system will be used to address various R&D projects that will have a positive impact on humankind, including:

• Predict the impact of global warming by simulating coastal or river flows directly linked to the Normandy region and its coasts or extreme atmospheric events.
• Accelerate innovation for autonomous driving by understanding obstacle avoidance across different environments, such as railways or adverse weather conditions.
• Advance healthcare initiatives through automated analysis of medical images, or the automatic annotation of sleep phases on electroencephalography signals.

Austral will be hosted at CRIANN in a customized HPE HPC pod, which will use direct liquid cooling (DLC) to optimize energy usage and reduce the overall carbon footprint of the system.

Austral will deliver more than two petaflops of peak performance, which is twice as much compute performance and ten times as much accelerated compute performance, compared to CRIANN’s previous resources.

Source Hewlett Packard Enterprise media announcement