Location: Istituto di BioRobotica – Scuola Superiore Sant’Anna
Lab head: Prof. Gastone Ciuti
Contact: marco.moscato@santannapisa.it
Lab Team
Gastone Ciuti
Professore Ordinario
Istituto di Biorobotica
Marco Moscato
Personale tecnico
Istituto di Biorobotica
Maria Mauro
Personale tecnico
Istituto di Biorobotica
Lab Mission
The SUPERCOMPLINK laboratory provides a high-performance computing node and a data storage system for large datasets. It supports research in machine learning, big data analysis, reinforcement learning and intensive model simulations (including, but not limited to multiscale neural models, biomechanical systems, biofluid dynamics systems and complex robotic platforms, i.e. soft/continuum robots)
Devices
Server
The system consists of two high-performance Dell servers (production node and development node) equipped with advanced hardware, details are listed below. These servers are designed to facilitate GPU-based computing and are actively used by researchers from various fields to enhance their computational workflows.
The production node has the following features:
- 2x Intel(R) Xeon(R) Gold 5215 CPU @ 2.50GHz (total of 20 cores and 40 threads)
- 92 GB RAM
- 2x Nvidia Quadro RTX 5000
- 4x HDD with 7.5 TB each
- 2x NVMe SSD with about 500 GB each
- 2x 10 Gigabit Ethernet ports
- 2x 1 Gigabit Ethernet ports
The development node has the following features:
- 2x Intel(R) Xeon(R) Gold 5217 CPU @ 3.00GHz (total of 16 cores and 32 threads)
- 124 GB RAM
- 3x Quadro RTX 6000
- 2x NVMe SSD with about 500 GB each
- 2x 10 Gigabit Ethernet ports
- 2x 1 Gigabit Ethernet ports
Data sheet: Nvidia Quadro RTX 5000; Xenon Gold 5215; Xenon Gold 5217
Others: Intel Guide
Note: Please contact the reference technicians for further information
AI calculation sytem
The AI computing system must consist of high-performance rack servers designed to support intensive workloads. Characteristics under acquisition (WIP)
Data sheet: download
Note:
Storage System
The storage solution shall be capable of storing large amounts of unstructured data. The system will be of the NAS Scale-Out type, capable of handling large files and supporting different network protocols for data access. At least 120TB of raw storage space, with at least one 800GB SSD disk per node used as a cache. Advanced data loss protection (fault tolerance on multiple disks or nodes) and protection against ransomware attacks will be provided.
Data sheet: download
Note:
ThinkStation PGX – Lenovo
The Lenovo ThinkStation PGX is a dedicated AI development workstation powered by the NVIDIA GB10 Grace Blackwell Superchip. It is designed to provide a localized, sandboxed environment for prototyping, fine-tuning, and inferencing AI models. The system supports diverse workloads, including machine learning, computer vision, generative AI/LLMs, robotics, and agentic design.
PERFORMANCE
Processor: Built into NVIDIA® GB10 Grace Blackwell Superchip, 20 core Arm, 10x Cortex-X925 + 10x Cortex-A725 Arm
Integrated Graphics: NVIDIA® Blackwell Architecture, Built into NVIDIA® GB10 Grace Blackwell Superchip
Chipset: NVIDIA® GB10 Grace Blackwell Superchip
Memory: 128 GB LPDDR5x, 256-bit, unified system memory
PCIe® NVMe® Drive: 4TB M.2 NVME SSD with self-encryption
Audio Chip: HDMI® multichannel audio output
Power Supply: 240W Adapter
Data sheet: download
Note: 2 devices available; just one user at a time on each device