About the HPC system Viper

June 01, 2024

Atos/Eviden system with AMD EPYC Genoa CPUs and AMD Instinct MI300A APUs

Supercomputer Viper is operational since June 2024 and comprises 768 compute nodes with AMD EPYC Genoa 9554 CPUs with 128 cores and at least 512 GB RAM per node. A subset of 609 nodes is equipped with 512 GB RAM, 90 nodes with 768 GB RAM, 66 nodes with 1024 GB RAM, and 3 nodes with 2304 GB RAM. Later in 2024, Viper will provide about 340 GPU compute nodes, each with 2 AMD Instinct MI300A APUs and 256 GB of high-bandwith memory (HBM3). The nodes are interconnected with a Nvidia/Mellanox NDR InfiniBand network using a fat-tree topology with two non-blocking islands, one for the CPU nodes (NDR200, 200 Gb/s), and one for the GPU nodes (NDR, 400 Gb/s). In addition there are 4 nodes for login and interactive tests, and I/O subsystems that serve ca. 20 PB of disk storage (attached to CPU and GPU system) with direct HSM access, plus ca. 500 TB of NVMe-based storage (attached to the GPU system).

Summary: 768 CPU compute nodes, 98304 CPU cores, 432 TB RAM (DDR5), 4.9 PFlop/s theoretical peak performance (FP64), ca. 340 GPU nodes comprising 680 APUs (to be deployed in the course of 2024).

Overall compute-node configuration:

609 CPU compute nodes (AMD EPYC Genoa 9554 with 128 cores per node and 512 GB RAM per node)
90 CPU compute nodes (AMD EPYC Genoa 9554 with 128 cores per node and 768 GB RAM per node)
66 CPU compute nodes (AMD EPYC Genoa 9554 with 128 cores per node and 1024 GB RAM per node)
3 CPU compute nodes (AMD EPYC Genoa 9554 with 128 cores per node and 2304 GB RAM per node)
ca. 340 APU compute nodes (2x AMD MI300A with 2x128 GB HBM3 per node) to be deployed in the course of 2024