Nvidia Accelerates Quantum Computing for Scientific Research

Nvidia has become synonymous with the current AI boom by supporting it with various chips and network platforms. However, the company is also focusing efforts on accelerating quantum computing, which Nvidia sees as having a synergistic relationship with traditional high-performance computing (HPC).

Nvidia-powered supercomputers are playing a role in the commercialization of quantum computing by supporting research led by Nobel laureate Giorgio Parisi and Massimo Bernaschi, director of technology at the National Research Council of Italy, who are focusing on quantum annealing. As outlined in research published in the July 2004 issue of the scientific journal Nature, it is a method that may one day tackle complex optimization problems that are extraordinarily challenging to conventional computers.

The researchers used more than two million GPU computing hours across multiple state-of-the-art facilities in Europe to simulate the behavior of a certain kind of quantum computing system known as a quantum annealer.

Quantum annealers are a special type of quantum computer that have limited use but may have advantages for solving certain types of optimization problems. They operate by methodically decreasing a magnetic field that is applied to a set of magnetically susceptible particles.

The paper published in Nature titled “The Quantum Transition of the Two-Dimensional Ising Spin Glass” addresses the problem of how the properties of magnetic particles arranged in a two-dimensional plane can abruptly change their behavior. The study also shows how GPU-powered systems play a key role in developing approaches to quantum computing by accelerating simulations that enable researchers to understand a complex system’s behavior in developing quantum computers.

The research comes on the heels of Nvidia announcing in May that it is providing its open-source Nvidia CUDA-Q platform to national supercomputing centers around the world, including sites in Germany, Japan and Poland. CUDA-Q is an open-source, quantum-classical accelerated supercomputing platform.

These sites are implementing the platform to power quantum processing units (QPUs)—which are the brains of quantum computers—inside their Nvidia-accelerated HPC systems. QPUs use the behavior of particles like electrons or photons to calculate differently than traditional processors, with the potential to make certain types of calculations faster.

Nvidia also announced that nine new supercomputers worldwide are using its Grace Hopper superchips to speed scientific research and discovery. Together, these supercomputing systems can deliver 200 exaflops, or 200 quintillion calculations per second, of energy-efficient AI processing power.

In Germany, the Jülich Supercomputing Centre (JSC) at Forschungszentrum Jülich is installing a QPU built by IQM Quantum Computers as a complement to its JUPITER supercomputer, which is running Nvidia’s GH200 Grace Hopper superchip. Meanwhile in Japan, the National Institute of Advanced Industrial Science and Technology is powering its ABCI-Q supercomputer with the Nvidia Hopper architecture to advance the country’s quantum computing initiative. Poland’s Poznan Supercomputing and Networking Center has recently installed two photonic QPUs connected to a supercomputer partition accelerated by Nvidia Hopper.

In a recent briefing, Dion Harris, head of data center product marketing at Nvidia, said that scientific computing spans a wide range of workloads and may include quantum computing.

Accelerated computing is a full stack challenge that starts with hardware, including CPUs, DPUs and GPUs that are put together to make a single system while layering on codex libraries and other domain specific acceleration libraries, Harris said. “Nvidia is invested and innovating at all layers of the stack.”

He said the systems using Grace Hopper are leveraging a novel architecture of tightly coupled CPUs and GPUs to support greater performance for HPC and AI, which allows for a more seamless interaction across these two simulations because of the fast connection between the CPU and GPU.

Tim Costa, who leads the HPC and quantum computing product team at Nvidia, said there is a tremendous opportunity to solve scientific challenges by integrating quantum processors into accelerated supercomputers.

Costa added that the relationship between quantum computing and traditional supercomputing will be analogous to the early days of GPU computing. “The supercomputer will handle most of the computation and it’ll hand off key parts to the quantum computer,” he said. “When we say quantum computer, we’re really talking about a processor. We’re not talking about a computer.”

There remain several challenges to realizing useful quantum accelerated supercomputing. “Today’s qubits are noisy and error prone,” Costa said. “Integration with HPC systems remains unaddressed.” Among the many other challenges are developing error correction, algorithms and infrastructure, he added.

The good news is there is a lot of investment being made in quantum, Costa said, with more than 25 national quantum initiatives launched across 350 quantum startups, and more than 48,000 quantum research papers have been published.

“The world has discovered that accelerated supercomputers are the key ingredient to address quantum challenges and realize this incredible opportunity as they’re both critical for quantum research, and they’re the systems in which quantum processors must be successfully integrated in order to be useful,” Costa said.