3 min read
In a significant leap for European scientific infrastructure, Paderborn University has launched its cutting edge supercomputer named Otus, developed in partnership with Lenovo and pro com Datensysteme GmbH. This system combines immense computational power with world class energy efficiency, marking a milestone in sustainable high performance computing (HPC).
Otus boasts approximately 142,656 processor cores, 108 GPU accelerators, and a memory pool of around 593.3 TiB. The system delivers a dual peak performance of 11.44 PFLOPS (CPU) and 7.45 PFLOPS (GPU) when measured by FP64 theoretical peak values. Its storage subsystem features around 5.4 PB of capacity using IBM Spectrum Scale technology. The interconnect fabric includes Nvidia NDR InfiniBand links with 200 Gbps and 2 × 400 Gbps connections depending on node type.
Otus is housed across 11 racks, utilising a mix of direct liquid cooling and air cooling variants. Its facility runs on 100 percent renewable electricity, and its hot exhaust heat is reused to warm campus buildings, aligning performance with sustainability. The system already ranks fifth globally on the Green500 list for energy efficient supercomputers.
High performance computing is increasingly essential across disciplines such as climate modelling, materials simulation, artificial intelligence, quantum computing, life sciences, and data analytics. By nearly doubling the computing capacity of its predecessor, Otus empowers researchers not only at Paderborn but also across Germany via the National High Performance Computing (NHR) Alliance.
The system’s sustainability credentials offer a model for how large scale computing can balance immense processing demands with environmental responsibility. As noted by Professor Christian Plessl of the Paderborn Centre for Parallel Computing:
With Otus we are developing an innovative and forward looking infrastructure … we are also setting standards in terms of energy efficiency and sustainability.
Paderborn University’s President, Professor Matthias Bauer, stated that the new infrastructure would enhance national research capabilities:
“High-performance computing is hugely relevant to the pressing challenges of our time,” he said. “Researchers can use supercomputers to trawl through huge quantities of data within a short period of time and identify patterns. Experiments that would be extremely expensive, dangerous or quite simply impossible can be simulated. This also allows us to make predictions about potential future developments… (and) access to state-of-the-art, world-class infrastructure.”
Otus serves as more than raw horsepower. Its architecture is designed for flexibility: it includes specialised large memory nodes (48 nodes each with 1,536 GiB RAM), GPU nodes with 2 × 96 core AMD EPYC Turin CPUs plus 4 × Nvidia H100 94GB GPUs, and FPGA nodes currently equipped with AMD Alveo V80 cards, with expansion planned up to 100 FPGA units.
Researchers will benefit from containerised workflows (Singularity), JupyterHub access, and a tailored data centre environment optimised for high reliability and high efficiency, including indirect hot water cooling systems.
Funding was supplied jointly by the federal government of Germany and the state of North Rhine Westphalia, with estimated total investment around €14 million (about US $15 million).
Though Otus is now operational for users (available since September 2025), its full impact will depend on how effectively researchers harness its capacity. Key questions include the system’s performance on complex AI and quantum workloads, its utilisation by external institutions, and how upgrades will be managed in a fast moving HPC landscape.
