GPU-based high-performance computing infrastructure

• Subject Area: Computer Science
• Term: since 2026
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 585120405

For researchers working on the use and new development of artificial intelligence (AI), GPU clusters have become an indispensable tool in recent years. It is no longer just computer science that is driving progress in this area. There is also growing demand from faculties such as electrical engineering, natural sciences, mechanical engineering, and process engineering. Therefore, a GPU cluster is to be procured that will be available to researchers from all OVGU institutes. Access to the GPU cluster is provided via the OVGU network infrastructure, so that the system is accessible and usable by the entire campus and thus by all scientists. There are three working groups that should be highlighted among the users from all parts of OVGU. The Artificial Intelligence (AI) Application in Production and Logistics (KIPUL) working group at the Institute for Engineering of Products and Systems (IEPS) is researching generative artificial intelligence (GenAI), in particular large language models (LLMs), as assistance tools for factory planning and logistics optimization. Current scientific research indicates that LLMs can analyze multimodal data in production, automatically generate simulation models, and effectively support the optimization of production strategies. The Artificial Intelligence Lab working group at the Institute for Intelligent Cooperative Systems (IKS) is working on applying basic knowledge of artificial intelligence in practice. Research here is motivated by real-world problems. Research focuses on deep learning methods, language, music, explainable AI, medical imaging, and education. The Chair of Software & Systems Engineering (CSSE) at IKS researches the interface between complex systems and secure, reliable software development. The working group develops intelligent software solutions that ensure the safety, reliability, and robustness of complex systems. Besides the two IKS working groups mentioned above, there are other working groups within the Faculty of Computer Science that depend on high GPU computing power to do their research. The proposed device configuration is based on the varied requirements of the users. The GPU cluster to be procured must be capable of efficiently computing large AI models. This will be achieved by powerful GPUs connected to SXM sockets with maximum bandwidth. The GPU cluster is configured so that users with lower requirements can also request individual GPUs. A development node is to be provided for software development.

• DFG Programme: Major Research Instrumentation
• Major Instrumentation: GPU-basierter Hochleistungsrechner
• Instrumentation Group: 7040 Vektorrechner
• Applicant Institution: Otto-von-Guericke-Universität Magdeburg
• Leader: Dr.-Ing. Stephan Jacob