Eviden Launches AI Computing Portfolio – High-Performance Computing News Analysis – insideHPC
Paris – May 2, 2024 – Eviden today announced the BullSequana AI product line that includes five AI server ranges for inferencing AI models at the edge, tuning existing models in enterprise data centers, and training foundation models with AI supercomputers.
Eviden said the portfolio leverages use case-focused AI software, including Eviden’s Computer Vision VISuite and global AI services provided by Eviden’s data scientist team. The products will also offer Nvidia AI Enterprise, a cloud-native software platform designed to accelerate data science pipelines and deployment of copilots and other generative AI applications.
The product’s first customer is GENCI (Grand Equipment National de Calcul Intensif) and CNRS (Centre National de la Recherche Scientifique), which have selected the system, powered by 1456 Nvidia H100 GPUs, to increase the peak computing power of their Jean Zay supercomputer from 36.85 to 125.9 petaflops. Operated by CNRS’s IDRIS (Institut de développement et de ressources en informatique scientifique), it will be available to users early this summer.
Eviden said the BullSequana AI 1200H is an enterprise-grade supercomputer designed for training and running inference of large language models (LLMs) and for complex systems modeling. The company said the system is both dense and energy efficient due to Eviden’s Direct Liquid Cooling (DLC) technology, enabling its AI supercomputers to be “at least three times denser than any traditional air-cooled systems, and additional heat can be transformed into energy to re-heat nearby buildings.”
The fiirst BullSequana AI 1200H DLC configuration is equipped with Nvidia Grace Hopper Superchips and Nvidia Quantum-2 InfiniBand networking. On the software side, it is equipped with Eviden Jarvice XE, Eviden Smart Energy Management Suite, Eviden Smart Management Center, and Nvidia AI Enterprise.
The system is manufacturered at Eviden’s factory in Angers, France, allowing businesses to meet their data sovereignty and security objectives, the company said.
Eviden is collaborating with Nvidia to integrate components in the portfolio, including Nvidia GH200 Grace Hopper Superchips. Eviden said it also plans to integrate NVIDIA GB200 Grace Blackwell Superchips.
“AI is a transforming force capable of revolutionizing businesses across industries and sectors by enhancing productivity and accelerating innovation,” said Dr. Cédric Bourrasset, head of Quantum & AI Computing, Eviden, Atos Group. “However, developing AI applications requires a great deal of expertise, high computing power, and energy to design, train, and run AI models. This is particularly true for foundation models. With Eviden’s new offering, customers who require reliability, scalability, and versatility to stay ahead of innovation will have maximized acceleration and the right expertise in AI to enable them to make their ideas and innovations a reality.”
“Eviden’s end-to-end solution, capitalizing on its expertise with the most renowned French and worldwide AI computing research centers, offers the most complete AI computing platform to help customers accelerate time to AI operationalization and industrialization of innovative applications in areas from healthcare, new energies, material design to legal or mass distribution” said Stephane Requana, CTO, GENCI.
This post was originally published on 3rd party site mentioned in the title of the post.