ORFEO

Modular HPC and AI data centre run by LADE, the Data Engineering Laboratory of Area Science Park. The operator's documentation lists 8 AMD EPYC 7H12 nodes, 13 Genoa nodes with two EPYC 9374F, 12 older Intel Xeon Skylake thin and fat nodes, 4 nodes with two V100 PCIe GPUs each, 2 DGX nodes with eight A100 SXM 40 GB GPUs each, and 3 Dell PowerEdge XE9680 nodes with two Xeon Platinum 8480+ and eight H100 SXM 80 GB GPUs each; summing that table gives 42 nodes, 2,856 CPU cores and 48 GPUs, a derived total that the page does not itself state, and the page carries no date and says four partitions while naming seven. The 22 October 2025 announcement describes a EUR 3 million upgrade from Italy's recovery plan (PNRR, Mission 4) adding three AI nodes of eight GPUs each (24 GPUs, model unnamed there but matching the three H100 nodes in the documentation), a new 125 kW server room with high-efficiency cooling, several more petabytes of storage and an SSD layer for hot data. Networking is InfiniBand up to 200 Gbit/s for RDMA (eight ConnectX-7 cards per H100 node), 25 Gbit/s Ethernet, and uplinks to the GARR and LightNet research networks. Start of service is not stated, so only the October 2025 evidence date is recorded; no FLOPS figure is published.