Australia’s Pawsey Supercomputing Research Center said today it will deploy eight of Nvidia Corp.’s powerful Grace Hopper Superchips nodes to power its open-source CUDA Quantum computing platform in Perth, as part of its mission to push the boundaries of quantum computing research.

The new supercomputer is expected to be able to deliver up to 10 times higher processing performance than the center was previously capable of, Nvidia said.

According to Pawsey, the Grace Hopper Superchips will enable researchers to run more advanced simulation tools and attempt to engineer new breakthroughs in areas such as quantum algorithm discovery, device design, machine learning, chemistry simulations, astronomy, image processing for radio, bioinformatics, financial analysis and other areas of research. Pawsey said the new chips will advance scientific exploration, both in Australia and in the wider world.

Nvidia’s Grace Hopper Superchips incorporate the company’s Grace central processing unit and Hopper graphics processing unit architectures and combine these with the Nvidia cuQuantum software development kit. Together, these elements form Pawsey’s open-source hybrid quantum computing platform, known as the CUDA Quantum platform.

The eight Grace Hopper Superchip nodes are based on Nvidia’s MGX modular architecture design, and are said to eliminate the need for a traditional CPU-to-GPU PCIe connection. They do this by “combining an Arm-based Nvidia Grace CPU with an Nvidia H100 Tensor Core GPU in the same package,” linking them via Nvidia’s NVLink-C2C chip interconnects.

One of the main benefits of the new interconnect is that it delivers more than seven times the bandwidth of the most advanced PCIe technology. As such, Pawsey’s researchers said they’re expecting a 10-fold increase in application performance when processing huge, multiterabyte datasets.

Nvidia said the Superchip nodes incorporate a single GH200 Grace CPU and one H100 GPU with 96 gigabytes of high-bandwidth memory. As such, the new installation at Pawsey will feature a total of eight Grace CPUs and eight H100 GPUs.

The advantage of the Nvidia CUDA Quantum platform is its hybrid design, which the company says is able to bridge the worlds of classical and quantum computing. According to Nvidia, it’s a first-of-its-kind platform that will enable “dynamic workflows across disparate system architectures.” The researchers intend to use the platform to integrate classical CPUs and GPUs with experimental quantum processing units in a single system that’s accelerated by yet more GPUs to enhance performance and scalability.

The new architecture will be able to run high-fidelity and scalable quantum simulations and seamlessly interface with future quantum hardware designs, Nvidia added.

Nvidia’s Director of HPC and Quantum Computing Tim Costa said high-performance simulation is necessary to address some of the biggest challenges in developing quantum computers, such as algorithm discovery, device design and methods for error correction, calibration and control. “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,” he said.

The new deployment will help to accelerate Australia’s status as a leader in the emerging quantum computing industry. The country is home to a number of startups working on some of the most compelling quantum computing designs, and its government has previously said the domestic market opportunity alone could be worth more than $2.5 billion per year. In addition, Australia hopes that by leading the quantum industry, it can create as many as 10,000 new jobs by 2040.

Image: Nvidia