Swiss Higher Technical School of Zurich (ETH Zurich) held the official launch ceremony of the Alps supercomputer at the Swiss National Supercomputing Center (CSCS) in Lugano. The system built by HPE has already taken the sixth place in the latest TOP500 rating and has an established FP64 performance of 270 Pflops (theoretical peak is 354 Pflops). By November, the remaining modules of the machine will be put into operation, and its maximum performance will be about 500 Pflops.

The June TOP500 ranking featured a section of 2,688 HPE Cray EX254n nodes with the "fantastic four" NVIDIA Quad GH200. To be more precise, this is still an "old" version of the accelerator with the H100 (96 GB HBM3), a 72-core Arm Grace processor and 128 GB LPDDR5x - a total of 10,752 Grace Hopper. This section consumes 5.2 MW and is in 14th place in the Green500. Nodes, of course, use the SOE.

This is the main, but not the only section of the supercomputer. Back in 2020, HPE deployed 1024 dual processor nodes with 64-core AMD EPYC 7742 (Rome) and 256/512 GB of RAM. Its performance is 4.7 Pflops. In addition, the Alps includes 144 nodes with one 64-core AMD EPYC, 128 GB of RAM and four NVIDIA A100 (80 or 96 GB HBM2e).

RAM and four AMD Instinct MI250X (128 GB HBM2e) and 128 nodes with four AMD Instinct MI300A hybrid accelerators. Most of the nodes will be connected by the HPE Slingshot-11: 200G interconnect-connection to a node or accelerator. A more precise configuration of the system will be revealed in November.

Lustre storage for the future machine was updated last year. The main storage is the Cray ClusterStor E1000 with a Slingshot-11 connection. Thus, 100 Pbytes of usable HDD capacity (8480 × 16 TB) with a bandwidth of 1 TB/s (300 thousand IOPS for writing, 1.5 million IOPS for reading) and 5 Pbytes of SSD, as well as backup capacities, were added. Two tape libraries with a volume of 130 PB each are responsible for archival storage.

A feature of the system is its geo-distribution (in fact, the nodes are located in four places) and a cloud-based usage model. Thus, the MeteoSwiss meteorological service of the country has received a dedicated virtual cluster at its disposal, which has already made it possible to switch to using a higher-resolution meteorological model that better reflects the difficult terrain of Switzerland. In addition, for safety reasons, some of the Alps nodes are located on the territory of the Federal Polytechnic School of Lausanne (EPFL).

Alps replaces the Piz Daint supercomputer (Cray XC50/40, 21.2 Pflops), the end of the life cycle of which was announced at the end of July 2024. For now, the CSCS will have the Arolla + Tsa machines (for the needs of MeteoSwiss) and Blue Brain 5 (solves the tasks of brain reconstruction and simulation). Alps, in addition to traditional HPC loads, will be used to develop AI solutions.