Leonardo - Computational Resources on GPU Supercomputer
Host Institution:
Consorzio Interuniversitario (CINECA), Italy
Direct applications to this installation are not possible. Computational budgets will be provided through HPC-based TAs according to the needs of specific projects.
Description:
Leonardo (booster partition) is the new operational accelerated system available since May 2023 in CINECA It is composed of more than 3456 compute nodes, each of them equipped with 32 cores Intel Xeon 8358 running at 2.6GHz and 4 NVIDIA Ampere A100 GPUs interconnected via Nvlink and having 64GB of HBM2 each. The internal network is Mellanox Infiniband HDR Dragonfly+ at 200 GB/s speed. The main focus of Leonardo is massively parallel computations for GPU-enabled applications. The cluster is actually number 4 of the Top500 list (TOP500 List - November 2022 | TOP500).
The service will be provided through trans-national access, which includes training, computational resources (with the support of CINECA help-desk for the correct usage of Leonardo facility). A total amount of more than 44 ML of core-hours will be available for TA and throughout the project.
More information:
Contact person:
Piero Lanucara (p.lanucaracineca.it)