Gadi Phase 2 — NCI is commissioning a new peak supercomputer to service the Australian computational research community for the coming years. NCI’s new supercomputer will be named Gadi (pronounced 'gar dee') after the words “to search for” in the language of the Ngunnawal people – the traditional owners of the Canberra region. When complete, Gadi will be a 3,200 node supercomputer from Fujitsu Australia.

Black computer servers with green highlights stacked up on top of each other in a dedicated enclosure.

The public launch is phased, with the first portion of the new machine launching in November and the second phase launching in January 2020. Phase 2 of Gadi will include more than half of Gadi's total CPU capacity, and a vast majority of its GPU capacity. While Phase 1 more than covers the computational performance and core count of the original Raijin Sandy Bridge cluster, Phase 2 will bring on board many more Cascade Lake CPUs and the latest V100 GPUs as well. The large-memory nodes, offering 1.5 Terabytes of Intel Optane DC Persistent memory, are also coming in Phase 2. For the latest information about the new machine, visit our dedicated information page.

We are also documenting the Gadi installation on our dedicated blog, click here to see the latest progress.

Technical Specifications

  • When complete, Gadi will include around 3,000 nodes containing Intel's second-generation Xeon Scalable ‘Cascade Lake’ processor with two 24-core CPUs and 192 Gigabytes of RAM per node.
  • Gadi will also include 160 nodes containing 640 Nvidia V100 GPUs, and 50 large-memory 'Cascade Lake' nodes offering 1.5 Terabytes of Intel Optane DC Persistent memory.
  • Linking the storage and the computer will be Mellanox Technologies' latest generation HDR InfiniBand technology in a Dragonfly+ topology, capable of transferring data at 200 Gb/s.
  • The underlying storage sub-systems will be provided by NetApp enterprise class storage arrays, linked together in a DDN Lustre parallel file system enabling the high-performance throughput needed for computing on big data challenges.
  • Altair’s PBSPro software will optimise job scheduling and workload management.
  • Gadi will use the latest version of the CentOS 8 operating system.