Will NVIDIA Enter The Race For Quantum Computing?

Well, not directly. However, the company is making it easier to develop code for quantum machines using GPUs.

Ok, let’s face it. Programming quantum computers is hard. REALLY hard. While we do not expect NVIDIA to develop and market their own quantum system any time soon (never say never!), the company is helping developers use GPU’s to simplify the coding task.

What has NVIDIA announced?

NVIDIA believes a “bridging” technology can help enable dynamic workflows across different architectures, with CPUs, GPUs, and quantum devices providing a hybrid quantum-classical computer platform. To enable this approach, NVIDIA has launched previews of the NVIDIA Quantum-Optimized Device Architecture (QODA), enabling programming in an integrated hybrid system and workflow.

For the programmer who wants to do algorithm research and build hybrid applications for future quantum advantage, a bridging technology is needed to enable dynamic workflows across disparate system architectures. The NVIDIA Quantum-Optimized Device Architecture (QODA) is a first-of-its-kind platform for hybrid quantum-classical computers, using CPUs, GPUs, and emulated QPUs to efficiently mimic how a hardware QPU would behave in a programmed situation.

“QODA consists of both a specification and a compiler NVQ++. It delivers a unified programming model designed for quantum processors (either actual or emulated) in a hybrid setting—that is, CPUs, GPUs, and QPUs working together,” said the company.

Quoda connects to any type of QPU backend, allowing accessibility to all users. Interestingly, NVIDIA has seen a 287X speedup in end-to-end Variational Quantum Eigensolver (VQE) performance with 20 qubits and dramatically improved scaling compared to existing Pythonic frameworks.

QODA features include:

  • Kernel-based programming model extending C++ for hybrid quantum-classical systems (full Python support is on the way).
  • Native support for GPU hybrid compute, enabling GPU pre- and post-processing and classical optimizations.
  • System-level compiler toolchain featuring split compilation with NVQ++ compiler for quantum kernels, lowering to Multi-Level Intermediate Representation (MLIR) and Quantum Intermediate Representation (QIR).
  • Standard library of quantum algorithmic primitives
  • Interoperable with partner QPUs as well as simulated QPUs using the cuQuantum GPU platform; partnering with QPU builders across many e different qubit types

Partners include several supercomputer centers including NERSC, Jülich, and Oak Ridge National Labs, and quantum pioneers Xanadu, Pasqal, Quantinuum, and others, however the big developers of QPUs (IBM, QCI, Google, Microsoft, Honeywell and others) seem missing at this time.

Conclusions

While it is unlikely that NVIDIA will build a quantum computer any time soon, if ever, the company’s GPUs and parallel software expertise can certainly help developers move forward to a time when a quantum computer can produce results that are superior to “traditional” (i.e., digital) computer systems. Quoda is another example of NVIDIA’s software and integrated systems approach that can deliver far more than its competitors who have only recently been able to match a 2 year old GPU in a few select benchmarks.