Expansion of Max Planck supercomputer Viper

The GPU-powered part of MPCDF’s supercomputer Viper was extended by another 27 nodes, thereby reaching its final expansion stage of 327 nodes with 654 AMD MI300A APUs.

The GPU-powered part of the new supercomputer Viper of the Max Planck Society, delivered by Bull (formerly Eviden/Atos), was extended by another 27 nodes, as a compensation for the delayed deployment of the entire Viper machine. With this extension Viper-GPU reaches its final configuration comprising 327 compute nodes (654 MI300A APUs).

The 327 Viper-GPU nodes are interconnected via InfiniBand NDR (400 Gb/s) and use IBM Storage Scale storage (HDD and NVMe) with approx. 12 petabytes capacity. Viper-GPU was opened for early-user operation in February 2025, formally accepted in December 2025 and has since then been running under full load. More details about Viper can be found on the MPCDF webpage and in the technical documentation for users.