High Performance Computing resources

Welcome to the local documentation for the high performance computing resources of the section for astrophysics and planetary science, and the neutrino astrophysics and biophysics groups at the Niels Bohr International Academy

Our groups have exclusive access to three clusters containing in total almost 6200 CPU cores and 8 very high performance GPUs. The clusters are complemented by a massive data storage archive and a number of powerful analysis machines used as frontends for the cluster and for pre- and post-prcessing.

The cluster is hosted at the High Performance Computing center at the University of Copenhagen.

The current cluster is the result of the collective effort of many groups with a diverse source of funding which has allowed us to keep the installation up to date. These include our 2020 installation of cascade-lake CPUs and Ampere GPU resources:

  • A Carlsberg Infrastructure grant to T. Haugbølle, KU funding to I. Tamborra, Novo grants to the Niels Bohr International Academy and U. G. Jørgensen, Villum Young Investigator grants to M. Bustamante, A. Doostmohammadi, J. Samsing and support from the section for astrophysics and planetary science made possible the acquisition in 2020 of 3360 CPU cores spread out on 70 nodes with 2x24 core Intel Cascade-Lake CPUs. Peak performance: 280 TFlops.
  • A Carlsberg Infrastructure grant to M. Pessah and P. Benitez-Llambay, Danish Council for Independent Research funding to T. Haugbølle and to J. K. Jørgensen, Novo and EU ITN grants to U. G. Jørgensen and support from the section for astrophysics and planetary science made possible the acquisition in 2020 of two GPU servers each configured with four Nvidia Ampere A100 GPUs, two AMD Epyc Rome CPUs, and 1 TB of memory. Peak performance: 80 TFlops. Machine learning FP16 Tensor core performance: 5 PFlops

The 2014 installation of ivy-bridge CPUs and Xeon-Phi accelerators was made possible by a Villum foundation grant to Å. Nordlund and a ERC starting grant to M. Pessah and included the acquisition of 139 ivy-bridge nodes with 78 Xeon-Phi knights corner accelerators. Peak performance: 62 TFlops (ivy-bridge) + 79TFlops (Xeon-Phi).
Additional nodes, 1 PB of storage, and four 32-core machines with 768GB of memory for analysis and post-processing have been acquired with support from the Centre for Star and Planet Formation, the Niels Bohr International Academy, and the Danish Council for Independent Research (four different grants by K. Galsgaard, O. Gressel, T. Haugbølle, and Å. Nordlund)

We are grateful to the Carlsberg Foundation, the Danish Council for Independent Research, the European Research Council, the Novo Fundation, and the Villum foundation for their continued support.