TEK2049

New Insights Into Supernova Powered by Cray EX System

Setonix supercomputer generates image of supernova remnant discovered in 1967

This image of the G261.9+5.5 supernova remnant, discovered back in 1964 by a CSIRO scientist, was generated by a brand new supercomputer, Setonix, based on the Cray EX platform. It is located at the Pawsey Centre, an Australian government-supported high-performance computing facility in Perth. The supernova is located somewhere between 10,000 and 15,000 light years away from Earth.

The enormous amount of data was collected by the ASKAP (Australian Square Kilometer Array Pathfinder) telescope, and transferred to the Pawsey Supercomputing Centre using a high-speed optical fiber link. The size, shape, and structure of the remnant revealed in the image should help scientists study the remnant and its surrounding medium in unprecedented detail.


PSC was founded in 2000, and over the years housed some of the world’s most powerful supercomputers. Past infrastructure included: Fornax (SGI), Zythos (SGI), Zeus (HPE).
Today however the facility supports three newer systems: Galaxy (Cray XC30), Magnus (Cray XC40), and the first stage of Setonix based on the HPE Cray EX platform.

In 2018 the center received another $70m to upgrade its supercomputing infrastructure. This allowed installation of the new AMD EPYC based system with a total of 65k cores. It also takes just one-fifth of the size compared with Magnus and Galaxy systems. The second stage of the Setonix system is planned to be completed by the end of 2022. The final upgrade should allow 30x more processing power than the older Cray systems.

System Technical Specifications

Setonix’s large, shared memory will allow us to use more of our software features and further enhance the quality of our images. This means we will be able to unearth more from the ASKAP data.

If you wonder what are the technical specifications of this system:

  • Over 500 AMD EPYC “Milan” CPU nodes (65k cores total).
  • 64 cores, at 2.55GHz, 2 per node.
  • 256GB RAM per node.

The system consists of several node types:

  • Eight 1 TB High Mem CPU nodes.
  • Eight Data mover nodes.
  • Sixteen Visualization nodes.
  • Four Login nodes.

The nodes are interconnected using HPE Slingshot 100Gb/sec connection.

Dr. Pascal Elahi said that processing data from ASKAP's astronomy surveys is a great way to stress-test the Setonix system and see what is possible. The data processing team is currently wrapping up a series of pilot surveys, and will soon undertake even larger and deeper surveys of the sky.