At its annual GTC event, Nvidia announced a partnership with Tel Aviv-based Quantum Machines to create a state-of-the-art architecture for quantum-classical computing.
The collaboration intends to bring about purpose-built infrastructure for quantum computing and GPU supercomputing capable of real-time quantum error correction. Known as DGX Quantum, the first system is expected to deploy to the Israel Quantum Computing Center.
>>Follow VentureBeat’s ongoing Nvidia GTC spring 2023 coverage<<
Why go for a hybrid quantum-classical architecture? It’s an effort to fill a capability gap while pure-bred quantum computing remains under construction. Quantum-classical bridges pair quantum algorithms or hardware with existing classical systems, so hybrid work can start here and now.
“If you’re trying to build the revolutionary computer [of the future], you still have to use the most revolutionary computer of the current time to create the ground truth to know whether the quantum computer is generating the right answers,” Jensen Huang, Nvidia’s CEO, said at a GTC press conference.
“You can’t just take algorithms developed for classical computing and think that’s going to be appropriate for quantum,” he said.
The quantum challenge
Today’s quantum computers have limited qubit-counts, and face serious error correction problems. But researchers continue to develop quantum algorithms that will exploit such computers when they eventually scale up.
Nvidia’s GPU-based DGX Quantum addresses these challenges. It matches an Nvidia Grace Hopper system with the CUDA Quantum open-source programming model and with the OPX quantum control platform from Quantum Machines.
The combination allows researchers to build applications that integrate quantum methods with cutting-edge classical computing, delivering calibration, control, quantum error correction and hybrid algorithms.
At GTC, Timothy Costa, Nvidia’s director of HPC and quantum computing products, cautioned that achieving quantum advantage is a difficult task, requiring solutions to numerous open challenges.
One challenge is performing error correction on hundreds of thousands to millions of qubits, which requires petascale computing and optimal latency between the compute and the QPU within the duration time of qubit coherence.
Another difficulty is that each qubit must be calibrated with numerous independent parameters needing optimization. That’s where DGX Quantum comes into play.
Itamar Sivan, CEO and cofounder of Quantum Machines, said that the DGX Quantum system has the potential to significantly reduce the barriers to integrated high-performance computing and quantum computing infrastructure. He predicts this integration will enable quantum infrastructure to scale faster and meet the increasing demand for quantum computing.
The states of quantum computing
The quantum-classical work anticipated at the Israel Quantum Computing Center betokens a trend of governments helping to sponsor quantum initiatives. At least 17 countries have invested in national programs for quantum technology research and development, according to a report by the World Economic Forum. Governments worldwide are making significant investments to support research institutes developing quantum computing technology. China, the U.S., Australia and countries of the European Union are among those investing in quantum computing initiatives.
The Canadian government recently announced a plan to invest at least $355 million (USD) in quantum talent, advancing the application of quantum technology and commercializing quantum computing as part of a new National Quantum Strategy. Similarly, the U.K. has announced a regulatory framework to support innovation in, and the ethical use of, quantum technologies.
Meanwhile, universities such as the Massachusetts Institute of Technology, Princeton University and the University of Waterloo are working collaboratively on developing quantum computer prototypes.
Nvidia as a quantum platform
A key player in quantum computing, Nvidia already boasts a long list of offerings that aim to accelerate quantum research, algorithm design, the development and discovery of applications, and tackling the challenge of building quantum integrated supercomputers — thereby taking the first steps to delivering on the promise of quantum computing at an industry level.
With the Nvidia quantum platform, researchers can simulate quantum processors at scale, and with performance far beyond what can be achieved on physical quantum processors today. This will enable them to design and develop better quantum algorithms for the processors of tomorrow. CUDA quantum developers can discover and test integrated quantum classical applications, using CPUs, GPUs, simulated QPS, and physical QPS together, each handling the parts of the workflow it does best.
And now, with the addition of DGX Quantum, customers can deploy tightly integrated quantum classical systems capable of using real-time GPU compute to make error correction, calibration control and hybrid algorithms possible at scale.
Until the quantum comes along
Nvidia’s Costa pointed out that the Nvidia quantum platform is focused on enabling and collaborating with the entire quantum computing-related economy, which has rapidly expanded over the past two years. Nvidia has partnered with quantum hardware builders, software companies and simulation frameworks, as well as system builders and integrators, major CSPs and research centers worldwide. Among Nvidia’s quantum partners beyond Quantum Machines are Atom Computing, IonQ and Oxford Quantum Circuits.
At GTC, Nvidia CEO Huang noted that “the quantum research community and development community is really vibrant around the world. There are a whole lot of interesting things to go and solve.”
Still, he cautioned that it is “solidly a decade and two decades away to have … broadly useful quantum systems.”
Even when those systems come along, quantum-classic hybrids will likely still be at work.