Eviden announces BullSequana AI 600, its new mid-range AI server family, powered by the AMD Instinct MI300X accelerator

Paris, France and Hamburg, Germany – May 13, 2024 – Eviden, the Atos Group business leading in advanced computing, today introduces the “BullSequana AI 600.” This new addition to the BullSequana AI product family, is launched in collaboration with AMD. The BullSequana AI 600 caters to businesses in need of 8-GPU servers to help developers, data scientists, and researchers achieve breakthroughs in terms of innovation and computation, with a special emphasis on models equipped with AMD Instinct™, the MI300X accelerator platform. It offers outstanding compute capabilities and a balance between cost and performance, ensuring that customers get the most value for their investment.

To assist enterprises and research in achieving their objectives with greater energy efficiency, the BullSequana AI 600 product family offers a comprehensive suite of AI systems. These servers are available in an air-cooled version (AI600) or can be equipped with Eviden’s Direct Liquid Cooling (DLC) technology for peak efficiency (AI600H). Eviden’s AI architecture, designed to cool both CPU and GPU boards, enables up to 80% heat retention through DLC. The DLC solution maintains density within the data center, leading to substantial cost savings for businesses and facilitating scalability.

A Strategic Alliance for Unmatched Performance-to-Cost Excellence

The BullSequana AI 600 utilizes AMD Instinct MI300X accelerators and the ROCm™ software suite which provides a robust and proven ecosystem for AI development, emphasizing openness, performance, and flexibility. Each BullSequana AI 600 is equipped with two 4^th Gen AMD EPYC™ processors and 8 AMD Instinct MI300X accelerators interconnected via an Infinity Fabric™ network, with each MI300X GPU providing 304 compute cores, 5.22 PFLOPS of theoretical peak performance (FP8), 2.61 theoretical peak performance PFLOPS (FP16) and 192 GB of HBM3 memory with 5.3 TB/s peak bandwidth.

Bruno Lecointe, Group VP Business HPC AI & Quantum, Eviden, Atos Group, said, “The collaboration between Eviden and AMD is a powerful technology partnership, yielding an exceptional performance-to-cost ratio. With the BullSequana AI 600, Eviden continues to deliver on its commitment to providing customers with the expertise and the high computing power needed to stay ahead on the AI innovation curve, while achieving their goals more energy-efficiently”.

Andrew Dieckmann, Corporate Vice President and General Manager, Data Center Accelerator Business, AMD, said, “Our collaboration with Eviden drives innovation for our customers in AI and high-performance computing. Powered by AMD Instinct MI300X accelerators and ROCm software suite, the new BullSequana AI 600 platform will help customers realize the performance and energy efficiency needed to address the rapid innovation in the AI market.”

50 systems in the TOP500, #1 in the Green500

Eviden now has 50 systems in the TOP500, the official listing of the world’s most powerful supercomputers, out of which 2 are in the TOP10. Entering the TOP500 for the first time, the CEA-HE extension ranks 17th in the TOP500, with a sustained performance of 57.11 petaflops for the extention. This accelerator partition is based on Eviden’s BullSequana XH3000 next-gen hybrid architecture. #1 in the Green500 is the first module of the exascale supercomputer JUPITER, named JEDI, from Forschungszentrum Jülich and EuroHPC Joint Undertaking. Based on the BullSequana XH3000, it was installed in April as a development instrument and prefigures the JUPITER exascale supercomputer which is currently being built at FZJ.