At ISC this week, Nvidia announced plans for a new hybrid classical-quantum computing lab with partners Jülich Supercomputing Centre and ParTec. The new lab is intended to run both high-performance and low-latency quantum-classical computing workloads. Nvidia also showcased work by Rolls-Royce and quantum software specialist Classiq to design and simulate the world’s largest quantum circuit for computational fluid dynamics (CFD) — a circuit that measures 10 million layers deep with 39 qubits.
Both the new lab and the Rolls-Royce work are significant progress points.
The new hybrid classical quantum lab will be built on the Jülich Unified Infrastructure (JUNIQ). Currently, JUNIQ is currently using the JUWELS booster system with 3,744 Nvidia A100 GPUs for quantum computing simulations. ParTec, which bills itself as “a pioneer in the field of Modular Supercomputing and Quantum Computing,” will work with Nvidia on integrating the lab with existing Jülich systems. Details of the planned new infrastructure were not disclosed.
Ian Buck, Nvidia vice president of HPC and hyperscale computing, said, “The Jülich supercomputing center is actually one of the world’s leading center in quantum research. They have already a quantum machine from D-Wave and they’re installing a Pasqal QPU in 2024. They may be one of the first centers to fully integrate and come up and with a classical supercomputer connected with quantum accelerator.
“We are partnering with them to help develop the programming model for these future devices. Of course, today, that’s done simulating on GPUs, but [it will be done by] integrating with actual QPUs in the future. The way people will program [in this lab] is with the CUDA quantum pro developer environment, where portions of the application can be specified to run on the QPU, not just on a GPU.”
Turning to the Rolls-Royce news, Buck said, “The announcement is that they’ve actually built a circuit that will help them move the computational fluid dynamics simulation of airflow and combustion to a quantum circuit. This is an extremely large circuit, over 10 million layers deep. It’s not a simple quantum experiment, but a working circuit mapped to their production CFD solver that they used to design their jet engines. While they can’t actually run this on a quantum computer today, they can be ready for those quantum computers.”
The Rolls-Royce work is impressive. Simulating a small number of qubits is relatively straightforward, but the number of possible states escalates quickly. Quantinuum researcher Henrik Dreyer recently told HPCwire that 32 qubits, for example, can have on the order of 4 billion states. European IT giant Atos/Eviden has long had a successful quantum simulator appliance – the quantum learning machine (QLM) – which uses Nvidia GPUs. The QLM was recently renamed Qaptiva 800 and will be part of the Eviden Qaptiva quantum portfolio after Eviden splits from Atos later this year. Nvidia launched its quantum development platform in March.
Rolls-Royce, of course, is a world leader in jet engine manufacturing, primarily for aircraft, but also power systems; it has been exploring quantum computing for some time. Among other challenges, it is hoped that quantum computers will help the European aviation industry reach a goal to reduce carbon emissions for jet engines by 75% by 2050.
“Applying both classical and quantum computing methods directly to the challenge of designing jet engines will help us accelerate our processes and perform more sophisticated calculations,” said Leigh Lapworth, computational science fellow at Rolls-Royce.
As noted in the announcement, “Rolls-Royce and its partner, Israel-based Classiq, designed the circuit using Classiq’s synthesis engine and then simulated it using Nvidia A100 Tensor Core GPUs. The speed and scale of the process was made possible by Nvidia cuQuantum, a software development kit that includes optimized libraries and tools to speed up quantum computing workflows.”
Nvidia has jumped into the quantum computing race full-force of late, but has no plans to build quantum computers.
“Nvidia is not building a quantum device. Instead, our focus is to help the world’s quantum engineers, researchers and developers simulate those future quantum machines. GPUs are ideal for simulating the superposition of quantum bits [and] the majority of quantum researchers today are simulating or using tools that are accelerated with Nvidia GPUs. Our platform [is] called Nvidia Quantum, which supports algorithms, research and also allows for building integrated quantum applications to simulate a future supercomputer, which has multiple kinds of accelerators, not just GPUs today, but potentially GPUs and QPUs along with CPUs,” said Buck.