HPE has announced its new artificial intelligence (AI)-based supercomputer, dubbed Champollion.
Designed in collaboration with Nvidia, the end-to-end supercomputer couples HPE’s machine learning development environment with Nvidia's accelerated computing platform.
The new AI supercomputer will also comprise 20 HPE Apollo 6500 Gen10 Plus systems, including 160 Nvidia Tensor Core GPUs for accelerated computing appropriate for AI model training and development, plus Nvidia Quantum InfiniBand for networking.
'Champollion' derives its name from Jean-François Champollion, a French linguist best known for deciphering Egyptian hieroglyphs. Per reports, the supercomputer will be hosted in HPE’s Center of Excellence (CoE) in Grenoble, France.
Commenting on upcoming plans to support virtual access to Champollion, HPE stated it intends to work on a flexible and demand-based system to help scientists and engineers expedite AI model testing. Champollion is currently available to a few select users. However, HPE affirms broader access will be available to the public in the near future so users can develop and train their models.
“As part of its global community access, Champollion will be used at 3IA research centers in Grenoble, Nice, Paris, and Toulouse, which were created in support of France’s AI mission,” confirmed HPE in a blog post.
“We are proud to continue our strong partnership with France’s leading organizations to support the nation’s AI vision. In addition to Champollion, we are building on existing HPE AI investments in France, such as the upcoming Adastra supercomputer for GENCI-CINES, the Jean Zay supercomputer for GENCI and a supercomputer for Maison de l'Intelligence Artificielle.”
Champollion marks yet another milestone for HPE, which has seen it win a string of high-profile supercomputing contracts over the past few years, including a deal to build the Czech Republic's most powerful high-performance computing installation in 2021.
In April 2021, HPE won a £22 million contract to develop a supercomputer for the National Supercomputing Centre in Singapore, considered to be eight times as powerful as its previous system. This was later followed by deals in Thailand, Hong Kong, and one to build Kestrel, a supercomputer ran by the Department of Energy’s National Renewable Energy Laboratory (NREL) in the US.
