March 25, 2020 — Pawsey Supercomputing Centre in Perth selects XENON Systems to deliver a new Linux based GPU cluster, named Topaz. This new cluster is the latest addition to Pawsey in its on-going program of expanding capabilities and delivering the latest technology to researchers. This refresh expands the capacity at Pawsey to meet the growing demands for GPU based processing. It provides the infrastructure for Pawsey researchers to continue to develop their skills on the latest GPU technology. This new cluster is complementary to the existing Zeus cluster, with Topaz providing researchers with enhanced visualization and GPU compute capabilities and increased batch processing speeds.
The new Topaz cluster includes the latest generation GPU and InfiniBand fabric to accelerate researchers’ work, in areas such as computational workloads, AI and machine learning, data analytics and data visualization. Topaz consists of 42 nodes: 20 nodes are for remote visualization and include the state-of-the-art NVIDIA Quadro RTX GPU’s, and 22 nodes are tuned for compute workloads and include dual NVIDIA Tesla V100 Tensor core GPU’s with 7 teraflops of double precision performance and more than 100 teraflops of deep learning performance per card. Across all the nodes, CPU workload is handled by dual Intel Cascade Lake processors. The theoretical double precision peak computational performance of Topaz exceeds 350 teraflops. Mellanox ConnectX-6 HDR InfiniBand at 200Gb/s optimizes the dataflow between the nodes and ensures researchers extract maximum performance from Topaz.
XENON Systems designed, delivered and installed the new system. XENON Systems pre-builds and extensively bench tests all customers’ HPC systems in their R&D facility in Melbourne, ensuring smooth delivery on-site with no issues. Topaz has been undergoing final configuration and testing with the assistance of a group of Pawsey researchers. Topaz went into full production as of early March 2020 and is available to projects with an active allocation on Magnus or Zeus supercomputer systems.
The latest state of the art CPU, GPU and InfiniBand fabric have delivered high performance, in a cluster, that will serve Pawsey researchers well into the future.
About Pawsey Supercomputing Centre
The Pawsey Supercomputing Centre is a world-class high-performance computing facility accelerating scientific discoveries for Australia’s researchers. Pawsey serves more than 1,800 active researchers from across Australia achieving unprecedented results, in domains such as radio astronomy, energy and resources, engineering, bioinformatics and health sciences.
The Pawsey Supercomputing Centre is an unincorporated joint venture between CSIRO – Australia’s national science agency – Curtin University, Edith Cowan University, Murdoch University and The University of Western Australia. It is supported by the Western Australian and Federal Governments. Learn more at www.pawsey.com.au
About XENON Systems
XENON Systems assists complex teams and organizations to do great new things. XENON has over two decades of experience in designing and deploying supercomputers, HPC clusters, and GPU based clusters. XENON also provides managed services for high performance computing environments, offering end-to-end, silicon to end user, support and management of high performance compute environments. XENON Systems is an NVIDIA Elite Solution Provider, and a Mellanox Premier Partner. Learn more at www.xenon.com.au
Source: XENON Systems