Infrastructure

HPC Cluster CALCULON

CALCULON is the central compute platform of our HPC Lab, designed to support a wide spectrum of scientific and AI workloads. The system combines a CPU cluster, operated in close collaboration with the University of Bern’s AIUB, with a modern GPU cluster tailored for data-intensive and accelerated computing. Its flagship nodes feature NVIDIA H200 GPUs interconnected via a high-speed 400 Gbit/s network, complemented by a diverse set of additional GPUs such as RTX A4500 and RTX 3080. This heterogeneous architecture allows users to match resources precisely to their workload requirements, from large-scale simulations to deep learning experiments.

The storage infrastructure follows a two-tier design to balance performance and capacity: a fast NVMe-backed shared filesystem available as /scratch for high-throughput workloads, and a large-scale HDD-based system for reliable long-term data storage. Beyond research, CALCULON plays an important role in teaching and student engagement: it supports student projects, provides GPU-backed JupyterHub environments for courses, and is used in educational initiatives such as the HPC summer school organized in collaboration with PSI. Together, these elements form a flexible and high-performance environment that evolves continuously with both research and education needs.

HPC Lab Space

The HPC Lab Space is the physical heart of our infrastructure—a large, versatile environment dedicated to experimentation, prototyping, and hands-on development. It is where new hardware is explored and assembled, from custom-built servers and cluster components to emerging technologies such as FPGAs and dataflow processors. The space encourages practical engagement with systems, enabling rapid iteration and deeper understanding of modern computing architectures.

Beyond infrastructure, the lab serves as a vibrant hub for education and collaboration. It hosts student projects, block modules, and provides a home base for students in the AI & HPC bachelor’s program. At the same time, it is where the CALCULON team works on extending and refining the cluster, ensuring that the platform remains at the forefront of performance, efficiency, and innovation.