HPE to build €250m AI exascale supercomputer in Germany
Hewlett Packard Enterprise is to build a next-generation AI supercomputer at the Leibniz Supercomputing Centre with 100% Direct Liquid Cooling.
The Blue Lion supercomputer will use HPE Cray technology and work with Nvidia Blackwell GPU accelerators and processors and come into operation in 2027. This will have 30x the performance of the current 26.7 PFLOPS SuperMUC-NG machine with 305,856 Intel 3.1 GHz Xeon Platinum 8174 cores. This would put the performance of over 800 PFLOPS, although with 8bit AI that performance is likely to hit the exascale range.
The supercomputer will use 100% direct liquid cooling to the processors with up to 40°C warm water, enabling the use of waste heat for the LRZ offices and could in future supply other organizations in the neighbourhood.
Blue Lion will be part of the German national HPC infrastructure of the Gauss Centre for Supercomputing and will serve a wide range of research projects, combining classic simulations with artificial intelligence for astrophysics, particle and quantum physics, fluid mechanics, natural sciences, life sciences and cultural sciences and many other research disciplines
The total cost of the machine and operation is €250m, and will be shared by the Federal Ministry of Education and Research (BMBF) and the Bavarian State Ministry of Science and the Arts (StMWK).
The system will also include fast data transfers between Blue Lion’s compute and storage units through HPE Slingshot, a high-performance interconnect that transfers 400 gigabits of data per second and allows jobs to scale across the entire system. LRZ will also gain purpose-built system management capabilities with HPE Performance Cluster Management, a software management tool that allows LRZ to efficiently monitor and manage the great scale of the supercomputer.
The interaction of GPU accelerators and CPU cores in high-performance computers such as Blue Lion requires a new way of programming software and algorithms. This is enabled by the HPE Cray Programming Environment, which helps migrate science code to the new system. To enable users to use the accelerated hardware of the supercomputer efficiently, LRZ and HPE will offer workshops and courses starting in 2025, helping researchers to optimize and port their applications. In addition, LRZ plans to closely cooperate with the European team from Hewlett Packard Labs, and LRZ will also increase its support team by 50 percent by the time Blue Lion goes into operation.
“Procuring a new supercomputer takes work, but it’s incredibly exciting. We can already take a look into the future of supercomputing. This increases the anticipation and even more the excitement of how the scientific community will use this system to make even better progress into new realms of knowledge. After all, it’s not about having the fastest supercomputer, but about providing the best possible support for cutting-edge research with our high-performance infrastructure,” said Prof. Dieter Kranzlmüller, Head of the Leibniz Supercomputing Centre.
“Supercomputing plays a pivotal role for progress in science and society, as well as for national competitiveness. LRZ’s Blue Lion is another big step to reinforce Germany’s position as a leading region for supercomputing and AI innovation. Blue Lion will significantly advance LRZ’s computing capacity and also position them in a cutting-edge position with next-generation technologies that will set new standards for the future of supercomputing,” said Heiko Meyer, Executive Vice President and Chief Sales Officer, Hewlett Packard Enterprise
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
