Scientific computing Technologies

Data servers

Our high-performance computing and data analytics platform (NEMO) is available to all Crick researchers.

Hardware:

  • High-performance research data storage with full offsite backup and long-term archive services. This has 22PB capacity as of January 2025, which continues to grow in line with new data requirements.
  • CPU Compute, including High Memory nodes. 12,500 physical compute cores with 2TB RAM standard nodes and 4TB high memory nodes.
  • GPU Compute.
  • Standard: 160 A100 (80GB SXM4) GPUs across 40 nodes.
  • Highest performance: 64 H100 (80GB SXM4) GPUs across 16 nodes.
  • Structural biology: 64 L40 GPUs across 16 nodes.
  • Networking: a high performance, low latency InfiniBand network is used to connect compute and storage nodes at speeds of up to 400Gbps.

Services:

  • Slurm scheduler-based batch computing using CPU and GPU.
  • Interactive nodes for CPU compute and Visualisation nodes (interactive GPU)
  • OnDemand – GUI web front end for those who do not wish to use ssh/command line. Includes MATLAB, Jupyter Notebook, RStudio Server, and image analysis tools as well as a Linux-based general desktop.
  • File Transfer including Globus, SFTP and to/from the major cloud providers
  • Support and provision of training.
  • We also have a GPU VM service for edge cases not covered by other services, such as Windows-based software.
  • We provide high-performance computing systems essential for capturing, storing and analysing scientific data.