HPC SYSTEMS

ENEAGRID is the infrastructure providing the access to the computational resources of ENEA ICT Division. Designed as a production quality, service oriented system for high performance and/or high throughput computing capabilities to support ENEA research activities. The infrastructure includes all the computational resources managed and developed by TERIN-ICT, a Division of the ENEA Energy Technology Department, connected in a Wide Area Network. The most relevant resources are the CRESCO Clusters based on x86_64 architecture. The main site is Portici ENEA Research Center hosting CRESCO8 (6 PFlops CPU, 3 PFlops GPU Pontevecchio), CRESCO7 (~500 TFlops CPU) and CRESCO6 (1.4 PFlops) clusters.

Other clusters, XCRESCO (ppc64le 1.9 PFlops GPU V100), CRESCO4F (~20 Tflops CPU) and CRESCO4C (10 Tflops CPU), are hosted on Frascati and Casaccia ENEA Research Centres. ENEAGRID provides a unified user environment and a homogeneous access method for all ENEA researchers, regardless of their location. It also implements tools facilitating the integration of departmental or individual resources, and supporting operation of experimental facilities.

The CRESCO infrastructure consists of multiple computational clusters designed to support a wide range of scientific and engineering workloads. Each cluster features a specific hardware configuration optimized for different computational requirements, including high-throughput processing, large-scale parallel simulations, data analysis, and AI applications.
The following sections provide an overview of the available clusters and their main hardware characteristics.

CRESCO8

CRESCO8 is the latest generation of ENEA’s High Performance Computing infrastructure and represents a significant advancement in computational capabilities available to the scientific community. Designed to support demanding research activities across multiple disciplines, the system provides a modern and scalable platform for large-scale simulations, data-intensive applications, and emerging AI workloads.

CRESCO8 Hardware Overview

  • 760 CPU nodes, each equipped with:
  • 2 × Intel® Xeon® Platinum 8592+ processors (64 cores per socket)
    Clock frequency ranging from 1.9 GHz to 3.8 GHz
    512 GB RAM
  • 17 GPU nodes, each equipped with:
  • 2 × Intel® Xeon® Platinum 8592+ processors (64 cores per socket)
    Clock frequency ranging from 1.9 GHz to 3.8 GHz
    512 GB RAM
    4 × Intel® Data Center GPU Max 1550 (Ponte Vecchio)
    128 Xe-Cores and 128 GB HBM2e memory per GPU
  • 15 HBM nodes, each equipped with:
  • 2 × Intel® Xeon® CPU Max 9480 processors (56 cores per socket)
    Clock frequency ranging from 1.9 GHz to 3.5 GHz
    1024 GB RAM
    128 GB High Bandwidth Memory (HBM)
  • Additional Features
  • Mellanox NDR 200 Gb/s high-speed interconnect
    25 Gb/s Ethernet interface per node
    BMC/IPMI support for remote management and console access
  • System Totals
  • 792 compute nodes
    101,136 CPU cores
    68 Intel Data Center GPUs (136 GPU tiles)
    High-bandwidth, low-latency Mellanox NDR 200 Gb/s network

TOP500

CRESCO8 is included in the TOP500 ranking (228 th) of the world’s most powerful supercomputers. This achievement reflects the system’s advanced computational capabilities and its ability to support large-scale scientific simulations, data-intensive workloads, and cutting-edge research activities across multiple disciplines. Check the information at this link.

Green500

In addition to its computational performance, CRESCO8 is also listed in the Green500 ranking (133 th), which evaluates the energy efficiency of high-performance computing systems. This recognition highlights ENEA’s commitment to developing sustainable HPC infrastructures through advanced cooling technologies and efficient resource utilization.

XCRESCO

XCRESCO is one of the main computing systems within ENEA’s High Performance Computing infrastructure. Hosted at the ENEA Research Center in Frascati, the cluster provides advanced computational resources that support multidisciplinary research projects, enabling efficient execution of complex simulations and data-intensive workloads.

XCRESCO Hardware Overview

  • Architecture
  • IBM Power AC922 (POWER9-based architecture)
  • 60 compute nodes, each equipped with:
  • 2 × IBM POWER9 processors
    16 cores per processor (32 cores per node)
    Clock frequency ranging from 2.6 GHz to 3.1 GHz
    Hyperthreading support (up to 4 threads per core)
    256 GB RAM (242 GB usable)
  • GPU Accelerators
  • 4 × NVIDIA Tesla V100 GPUs per node
    NVLink 2.0 high-speed interconnect
    16 GB memory per GPU
  • Additional Features
  • Mellanox InfiniBand EDR 100 Gb/s network
    140 TB usable NFS storage
    Distributed across 3 racks
  • System Totals
  • 60 compute nodes
    1,920 CPU cores
    240 NVIDIA Tesla V100 GPUs
    Approximately 1.9 PFLOPS peak performance
    Up to 32 TFLOPS per node
    Mellanox InfiniBand EDR 100 Gb/s interconnect

CRESCO7

Located at the ENEA Research Center in Portici, CRESCO7 is a key component of the CRESCO High Performance Computing infrastructure. The system provides a robust and reliable computing environment that supports a wide range of scientific applications, including numerical simulations, data analysis, and multidisciplinary research activities.
Designed to deliver high computational performance and scalability, CRESCO7 continues to serve as an important resource for researchers and projects requiring advanced HPC capabilities.

CRESCO7 Hardware Overview

  • Architecture
  • Intel Xeon Platinum-based architecture
  • 144 compute nodes, each equipped with:
  • 2 × Intel® Xeon® Platinum 8160 processors
    24 cores per processor (48 cores per node)
    Clock frequency of 2.10 GHz
    192 GB RAM
  • Network
  • Mellanox EDR high-speed interconnect
    100 Gb/s low-latency network
  • Additional Features
  • Dual Gigabit Ethernet interfaces per node
    BMC/IPMI support for remote management and console access
  • System Totals
  • 144 compute nodes
    6,912 CPU cores
    Mellanox EDR 100 Gb/s interconnect

CRESCO4C

Located at the ENEA Research Center in Casaccia, CRESCO4C is a 32-node computing cluster and an integral part of the CRESCO High Performance Computing infrastructure. The system provides computing resources for scientific applications, including numerical simulations, parallel processing, and data analysis.

CRESCO4C Hardware Overview

  • Architecture
  • Intel Xeon E5-2670-based architecture
  • 32 compute nodes, each equipped with:
  • 2 × Intel® Xeon® E5-2670 processors
    8 cores per processor, for a total of 16 cores per node
    Clock frequency of 2.60 GHz
    64 GB RAM, corresponding to 4 GB per core
  • Network
  • InfiniBand 4× QDR high-bandwidth, low-latency interconnect
    40 Gb/s nominal bandwidth
    32 Gb/s effective data transfer rate
    QLogic 12800 network switch
    2 × Gigabit Ethernet interfaces per node
  • Additional Features
  • BMC/IPMI 1.8 support
    Remote console management software
  • System Totals
  • 32 compute nodes
    512 CPU cores
    2 TB total RAM
    16 TB aggregate local storage
    InfiniBand 4× QDR 40 Gb/s interconnect

CRESCO4F

Located at the ENEA Research Center in Frascati, CRESCO4F is a 64-node computing cluster and an integral part of the CRESCO High Performance Computing infrastructure. The system provides a reliable computing environment for scientific applications, including numerical simulations, parallel processing, and data analysis.

CRESCO4F Hardware Overview

  • Architecture
  • Intel Xeon E5-2670-based architecture
  • 64 compute nodes, each equipped with:
  • 2 × Intel® Xeon® E5-2670 processors
    8 cores per processor, for a total of 16 cores per node
    Clock frequency of 2.60 GHz
    64 GB RAM, corresponding to 4 GB per core
  • Network
  • InfiniBand 4× QDR low-latency interconnect
    40 Gb/s nominal bandwidth
    32 Gb/s effective data transfer rate
    QLogic 12800 network switch
    2 × Gigabit Ethernet interfaces per node
  • Additional Features
  • BMC/IPMI 1.8 support
    Remote console management software
    GPFS parallel file system
  • System Totals
  • 64 compute nodes
    1,024 CPU cores
    4 TB total RAM
    32 TB aggregate local storage
    InfiniBand 4× QDR 40 Gb/s interconnect

The CRESCO infrastructure has evolved over time through several generations of computational clusters, some of which have now been decommissioned. These systems supported a wide range of scientific and engineering activities, contributing to high-performance computing, parallel simulations, data processing, and research applications throughout their operational lifetime. The following sections provide an overview of the decommissioned clusters and their main hardware characteristics.

CRESCO5F

The CRESCO5F cluster was located in Frascati.

CRESCO5F Hardware Overview

  • 8 nodes, each equipped with:
  • 2 × AMD EPYC 7313 processors (16 cores per socket)
    Clock frequency of 3.0 GHz
  • 6 NVIDIA® A100 GPUs
    256 GB RAM (8 GB per core)
    1 × 500 GB SATA II disk
    InfiniBand MLX5 100 Gb/s interface (approximately 80 Gb/s effective bandwidth)
    2 × 10 GbE interfaces
  • Additional Features
  • BMC/IPMI support and software for remote management and console access

CRESCO5

The CRESCO5 cluster was located in Portici.

CRESCO5 Hardware Overview

  • 40 CPU nodes, each equipped with:
  • 2 × Intel® Xeon® E5-2630 v3 processors (8 cores per socket)
    Clock frequency of 2.40 GHz
    64 GB RAM (4 GB per core)
    Up to 16 instructions per clock cycle
    InfiniBand QDR interface (40 Gb/s, approximately 32 Gb/s effective bandwidth)
    2 × 1 GbE network interfaces
  • Additional Features
  • BMC/IPMI 1.8 support and software for remote management and console access

CRESCO4

The CRESCO4 cluster was located in Portici.

CRESCO4 Hardware Overview

  • 304 CPU nodes, each equipped with:
  • 2 × Intel® Xeon® E5-2670 processors (8 cores per socket)
    Clock frequency of 2.60 GHz
    64 GB RAM (4 GB per core)
    Up to 8 instructions per clock cycle
    1 × 500 GB SATA II disk
    InfiniBand QDR interface (40 Gb/s, approximately 32 Gb/s effective bandwidth)
    2 × 1 GbE network interfaces
  • Additional Features
  • BMC/IPMI 1.8 support and software for remote management and console access

CRESCO3

The CRESCO3 cluster was located in Portici.

CRESCO3 Hardware Overview

  • 84 CPU nodes, each equipped with:
  • 2 × AMD® Opteron® 6234 processors (12 cores per socket)
    Clock frequency of 2.40 GHz
    64 GB RAM
    1 × 500 GB SATA II disk
    InfiniBand QDR interface
    2 × 1 GbE network interfaces
    1 × video card
  • Additional Features
  • BMC/IPMI 2.0 support and software for remote management and console access

CRESCO2

The CRESCO2 cluster was located in Portici.

CRESCO2 Hardware Overview

  • 56 Nehalem CPU nodes, each equipped with:
  • IBM HS21 blade server
    2 × Intel® Xeon® E5530 processors (4 cores per socket)
    Clock frequency of 2.40 GHz
    8 MB L3 cache per processor
    16 GB RAM
    InfiniBand network interface
  • Additional Features
  • InfiniBand high-speed interconnect
    Integration with the ENEA HPC infrastructure

CRESCO1

The CRESCO1 cluster was located in Portici.

CRESCO1 Hardware Overview

  • 42 CPU nodes, each equipped with:
  • IBM System x3850-M2 server
    4 × Intel® Xeon® E7330 processors (4 cores per socket)
    Clock frequency of 2.40 GHz
    32 or 64 GB RAM
    InfiniBand network interface
  • Additional Features
  • Integration with storage and service nodes within the ENEA HPC infrastructure
    Decommissioned on November 18, 2014
    Compute nodes subsequently reused in the DTE-ICT division’s cloud infrastructure