Fork me on GitHub

JADE II (Tier 2 GPU cluster)

The JADE II cluster is a 2020 renewal of the Joint Academic Data Science Endeavour (JADE) maintaing its status as the largest GPU facility in the UK supporting world-leading research in machine learning.

The computational hub harnesses the capabilities of the NVIDIA DGX MAX-Q Deep Learning System and comprise of 63 servers, each containing 8 NVIDIA Tesla V100 GPUs linked by NVIDIA’s NV link interconnect technology. The MAX-Q range are a more power-efficient system allowing the doubling of computational power of JADE with only 2/3 of power that would have been required.

The RSE group director (Paul Richmond) is a co-investigator in the EPSRC Tier 2 JADE II system. As such members of the University of Sheffield are able to access this resource for free (for use with deep learning research).

Note

The system is expected to be launched and accessible to researchers by the end of 2020.

JADE II Specification

More details on the system coming soon.

Hardware

  • 63 Nodes of DGX MAX-Q, each with:
    • 8x Nvidia V100 32GB GPUs

    • 512GB RAM

  • 70TB DDN AI400 shared storage (NVMe) for read intensive/streaming applications

  • 1PB Lustre shared storage (spinning disk)

  • EDR infiniband interconnect

Software

  • Redhat Linux