High performance computing for AI research

New infrastructure launched for deep learning projects – services offered to research institutions in Bavaria 

AI researchers at all Bavarian research universities and universities of applied sciences can now use the new IT infrastructure provided by Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU) for their projects. The Erlangen National High-Performance Computing Center (NHR@FAU) has installed several hundred AI accelerators financed by the Free State of Bavaria within the framework of the High-Tech Agenda Bavaria. NHR@FAU not only provides the computing power for deep learning projects, but also offers comprehensive services ranging from helping users choose resources or organize their workflow and providing courses in how to use the system.

“At the current time, the two clusters ‘Alex’ and ‘Helma’ are installed at NHR@FAU. Their hardware is suited to AI research and also offers the necessary computing power for other demanding simulation projects from various research areas,” explains Prof. Dr. Gerhard Wellein, head of NHR@FAU. Alex is already in operation and is equipped with more than 600 graphics accelerators, known as GPUs, from the market leader Nvidia. The Helma cluster was installed in fall 2024, and initially equipped with 384 Nvidia GPUs of the latest generation (H100). Selected groups of users are currently putting Helma to the test.

Researchers at research institutes across Bavaria can access the services offered by NHR@FAU straightforwardly and free of charge via the “BayernKI” portal (ki.bayern.de/wissenschaft). “As academic research into AI is progressing at a breakneck speed, we deliberately decided to make it easy for researchers to access our services,” explains Gerhard Wellein. “Anyone who has a good justification for using the services can get started in a matter of days.”

First projects already underway

The new AI infrastructure is already being used extensively. The Department of Artificial Intelligence in Biomedical Engineering (AIBE) at FAU, for example, is investigating how AI can generate ultrasound images of the heart. These artificially created but realistic imaging data are not subject to any data protection restrictions and can be used to train other AI models or medical professionals. Another project located at the Chair of Pattern Recognition at FAU is developing a language model for German tax law that creates high quality training material using an automated data pipeline. The aim is to make AI-supported analyses of complex areas of law publicly available.

For the AI projects, NHR@FAU does not only offer the required computing power but also comprehensive services: Users receive individual support for selecting the most suitable hardware resources, adapting software, optimizing their programs, managing their data and organizing their workflow in general. Furthermore, staff at NHR@FAU monitor the ongoing computing jobs and regularly offer workshops and courses for beginners and advanced users.

Further expansion planned for 2030

The Free State of Bavaria is providing a total of 55 million euros for expanding high performance computing capacities in the area of AI at NHR@FAU and at the Leibniz Supercomputing Center in Garching (LRZ). This funding will be used to expand AI resources for researchers in Bavaria at both locations over the coming years. However, as a national center, NHR@FAU also provides capacities for researchers across the whole of Germany. As the massive expansion of hightech infrastructure requires a lot of space and a modern environment, the high performance computers will be relocated as of 2030 together with NHR@FAU and parts of the Erlangen Regional Computing Center (RRZE) to a new, energy-efficient and sustainable computing center. The Free State of Bavaria is providing up to 270 million euros in funding for the construction project at FAU Campus Erlangen (South).

Further information:

Prof. Dr. Gerhard Wellein
Phone: +49 9131 85 28136
gerhard.wellein@fau.de