Nvidia teases quantum accelerated supercomputers

At ISC High Performance 2024 in Hamburg, Germany, Nvidia today announced that nine new supercomputers worldwide are using its Grace Hopper Superchips to deliver a combined 200 exaflops (200 quintillion calculations per second) of computing power with, it said, twice the energy efficiency of an x86 system plus GPU.

Grace Hopper accounts for 80% of Hopper sales, said Dion Harris, Nvidia’s director, accelerated data center GTM, during a media briefing. “The reason why that’s exciting is that it leverages this novel sort of architecture of this tightly coupled CPU and GPU architecture to deliver great performance for HPC and AI.”

The first European Grace Hopper supercomputer to come online is Alps at the Swiss National Supercomputing Centre, which was built by Hewlett Packard Enterprise (HPE) and offers 20 exaflops of AI computing driven by 10,000 Grace Hopper superchips. Its role is to advance weather and climate modeling, and material science.

Nvidia also announced that national supercomputing centers worldwide will soon receive a performance boost via the open-source Nvidia CUDA-Q platform. The company revealed that sites in Germany, Japan, and Poland will use the platform to power quantum processing units (QPU) in their high performance computing systems.

“Quantum accelerated supercomputing, in which quantum processors are integrated into accelerated supercomputers, represents a tremendous opportunity to solve scientific challenges that may otherwise be out of reach,” said Tim Costa, director, Quantum and HPC at Nvidia. “But there are a number of challenges between us, today, and useful quantum accelerated supercomputing. Today’s qubits are noisy and error prone. Integration with HPC systems remains unaddressed. Error correction algorithms and infrastructure need to be developed. And algorithms with exponential speed up actually need to be invented, among many other challenges.”

To address these issues, he said, more than 25 national quantum initiatives have been launched. There are more than 350 quantum startups, over 70% of the Fortune 500 have some sort of quantum program, and more than 48,000 quantum research papers have been published.

“But another open frontier in quantum remains,” Costa said. “And that’s the deployment of quantum accelerated supercomputers – accelerated supercomputers that integrate a quantum processor to perform certain tasks that are best suited to quantum in collaboration with and supported by AI supercomputing. We’re really excited to announce today the world’s first quantum accelerated supercomputers.”

These machines will be at AIFST in Japan, Jülich in Germany, and PSNC in Poland (which has installed two QPUs).

“The integration of not one but four quantum processing units with three supercomputers opens the door to the next wave of quantum innovation,” said Heather West, research manager, quantum computing, infrastructure systems, platforms, and technology group, at IDC. “Researchers have always expected that quantum computing would accelerate scientific advantage. However, the symbiotic relationship between quantum-classical compute technologies will also help to accelerate the development of quantum systems themselves, paving the way for useful, error-corrected, quantum-centric supercomputers and the era of quantum utility, a long awaited destination for both quantum researchers and quantum end users.”

However, said Harris, this will need the application of AI models to succeed. “We  don’t think that there will be a successfully deployed fault-tolerant system that doesn’t use AI models to do large scale, real-time error correction to calibrate these devices. Right now, it’s an incredibly human-time-intensive task for physicists to calibrate and keep up a quantum device. And it’s only going to get harder and harder as the number of qubits goes up. And so we have to automate that and apply the best technology and AI in order to do those tasks.”