Nvidia’s QODA, cuQuantum hybrid compute frameworks gaining in adoption

Nvidia’s QODA (Quantum Optimized Device Architecture) platform, announced in July, is designed to provide a single development environment for applications that use a hybrid of quantum and classical (i.e., conventional, non-quantum) code. Today, at its annual GTC (GPU Technology Conference) developer event, Nvidia announced that adoption of QODA and a companion project called cuQuantum, have taken off.

Check out: Nvidia reveals QODA platform for quantum, classical computing

Since quantum computers are still in early stages of development, they’re in short supply, and expensive. It also turns out, though, that GPU platforms, though classical, are good ones for emulating quantum circuits. That means they can execute both quantum and classical code on the same physical platform, making QODA’s vision a natural one for Nvidia to pursue. The mainstream environment for this scenario, by the way, is in HPC (high-performance computing) setups in the cloud, where Nvidia GPUs are customary.

[Follow along with VB’s ongoing Nvidia GTC 2022 coverage »]

QODA’s uptake

At Microsoft, the Azure Quantum team, working with Quantum startups Quantinuum, Quantum Circuits Inc., and Rigetti Computing, along with Oak Ridge National Lab, are using QODA in their work to demonstrate how classical and quantum systems can be hybridized to take on big compute requirements.

Bettina Heim, a Microsoft principal software engineering manager on the team, who also has a Ph.D. in quantum computing, says “QODA will help us take that next step of offering a robust integration of very different systems and developer tools.” Heim also chairs the QIR Alliance, an industry group focused on quantum, and it’s using QODA too.

Nvidia reports a number of other instances of QODA adoption in the quantum computing ecosystem. Outside of the Azure Quantum work, Rigetti Computing will let developers use QODA on its superconducting quantum computers; Classiq will use QODA to optimize performance of its quantum algorithms; and IQM Quantum Computers, Pasqal, Quantinuum, Quantum Brilliance and Xanadu all announced QODA initiatives when the framework was announced in July.

cuQuantum simulations popular, too

cuQuantum is a companion SDK project to QODA, specifically focused on quantum circuit simulations on GPUs. Nvidia says developers can create accurate simulations of hundreds of qubits (quantum bits) on a single NVIDIA A100 Tensor Core GPU with cuQuantum. Moreover, Nvidia says it’s possible to create the equivalent of thousands of qubits on a GPU-based supercomputing cluster. Since some of the best physical quantum machines max out at a few hundred qubits right now, that’s serious quantum business.

cuQuantum is picking up adoption too. BMW Group, collaborating with Amazon Web Services (AWS), is using cuQuantum. So is Japan’s Fujifilm Informatics Research Laboratory, in collaboration with blueqat. Oracle is making the DGX cuQuantum Appliance (a software container with all the components needed to run cuQuantum jobs) available on the Oracle Cloud Marketplace. Google Quantum AI and IonQ had already announced support for cuQuantum last year.

AWS announced the availability of cuQuantum on its Braket service, and recently demonstrated how the framework can provide a 900-times speedup on quantum machine learning (ML) jobs. In a blog post, AWS Braket research scientist Yi-Ting (Tim) Chen, said, “By switching from the CPU simulator in a single instance to parallelized training with the GPU simulator, we improve the run time by about 900x and reduce the cost by about 3.5x.”

Deloitte is using QODA and cuQuantum together for customer service natural language processing (NLP) applications and Nvidia says Deloitte will explore putting the tech to work in drug discovery scenarios, too. The R&D group at SoftServe is also exploring the use of cuQuantum in drug discovery, as well as in the optimization of emergency logistics.

Quantum eagerness

This is a lot of adoption, and it’s quick. cuQuantum was announced less than a year ago and QODA was announced less than two months ago. It’s clear the research community wants to put quantum to work on a number of hard problems, including ones that can save lives. Quantum hardware is still incubating, but GPU hardware lets research teams do quantum work now. Nvidia’s wasting no time making things work on its GPUs, and pitching it all to developers at GTC should accelerate things further.