Australia’s Pawsey Supercomputing Research Centre is the latest high-performance computing hub to adopt Nvidia’s Grace Hopper Superchips, as well as the semiconductor giant’s will add the CUDA Quantum platform and related software in a move that will showcase the potential of hybrid classical-quantum computing on a massive scale.
The Pawsey plan, announced at this week’s SupercomputingAsia 2024 event in Sydney, calls for the deployment of eight Grace Hopper GH200 Superchip nodes connected in a modular architecture at its National Supercomputing and Quantum Computing Innovation Hub in Perth. Pawsey researchers also will leverage CUDA Quantum, an open-source hybrid quantum computing platform that supports high-performance simulation, and features capabilities to program hybrid CPU, GPU and QPU (quantum processing unit) systems. They also will take advantage of Nvidia’s cuQuantum software development kit of optimized libraries and tools for accelerating quantum computing workflows.
The Pawsey hub has used Nvidia GPUs for several years. Now, it will get a platform with GH200 Superchips based on the Nvidia MGX modular architecture. GH200 Superchips, which also are being deployed at similar sites, such as Germany’s Jupiter supercomputing facility, combine an Arm-based Nvidia Grace CPU with an Nvidia H100 Tensor Core GPU in the same package, using the company’s NVLink-C2C chip interconnects. The result is a 7x bandwidth increase between the CPU and GPU when compared with the latest PCIe technology, Nvidia claims. It also delivers, the company says, up to 10x higher performance for applications running terabytes of data.
That will come in handy for Pawsey, an institution which over the last two years has advanced to the cutting-edge of hybrid classical-quantum supercomputing by deploying rack-mounted quantum computing equipment from Australia’s Quantum Brilliance at the same site as Pawsey’s traditional classical high-performance computing gear in a bid to merge and maximize the power of both systems. It is worth noting that Nvidia announced a partnership with Quantum Brilliance almost a year ago, and has been a proponent of “GPU-accelerated” quantum computing. Adding GH200 Superchips and quantum software to the mix will enable Pawsey to conduct highly complex simulations on a massive scale.
“High-performance simulation is essential for researchers to address the biggest challenges in quantum computing, from algorithm discovery and device design to the invention of powerful methods for error correction, calibration and control,” said Tim Costa, director of HPC and quantum computing at Nvidia. “CUDA Quantum, together with the Nvidia Grace Hopper Superchip, allows innovators such as Pawsey Supercomputing Research Centre to achieve these essential breakthroughs and accelerate the timeline to useful quantum-integrated supercomputing.”
Mark Stickells, executive director at the Pawsey Supercomputing Research Centre, added, “Pawsey Supercomputing Centre’s research and test-bed facility is helping to advance scientific exploration for all of Australia as well as the world. Nvidia’s CUDA Quantum platform will allow our scientists to push the boundaries of what’s possible in quantum computing research.”