Research Computing Services HPC Documentation¶
The Cambridge Service for Data Driven Discovery (CSD3) is one of the EPSRC Tier2 National HPC Services hosted by Research Computing Services at the University of Cambridge.
The facility became live in November 2017 and began the next chapter in HPC for the University of Cambridge. It was upgraded in 2020, and again in 2021 and 2023.
CSD3 is currently a combination of two platform types:
- Intel Xeon Scalable Processors (Cascade Lake, Ice Lake and Sapphire Rapids CPUs)
- NVIDIA A100 GPUs.
All CSD3 nodes are currently interconnected by Mellanox HDR InfiniBand. The GPU sub-cluster is also known as Wilkes3-GPU in its own right. The name “Peta4” was associated historically with the original CPU elements of CSD3.
CSD3 is available to academic researchers both internal and external to Cambridge, and to industry. For more information please see the website.
Getting Started
Dawn/AIRR/Zenith
CSD3 User Guide
- Connecting to CSD3
- Login-Web Interface
- Connecting to CSD3 via TurboVNC (3D Visualisation)
- Running Jobs on CSD3
- Cascade Lake Nodes
- Ice Lake Nodes
- Ampere GPU Nodes
- Sapphire Rapid Nodes
- Sapphire Rapid Nodes with High Bandwidth Memory
- WBIC-HPHI Platform End-of-Life
- End of SBS allocation scheme
- Long jobs
- Interactive use
- Modules Environment
- Using Spack
- Using Python
- Performance Tips
- File and I/O Management
- Data Transfer Guide