Nvidia is collaborating with Google Quantum AI to speed up the design of its next-generation quantum computing devices using Nvidia simulations.
Google Quantum AI uses the hybrid quantum classical computing platform and the Nvidia Eos supercomputer to simulate the physics of its quantum processors. This will help overcome the present limitations of quantum computing hardware, which might only perform a certain variety of quantum operations before computations need to stop attributable to what researchers call “noise.”
“Developing commercially viable quantum computers is just possible if we are able to scale
“Building quantum hardware while keeping noise under control,” said Guifre Vidal, research scientist at
Google Quantum AI, in a press release. “Using Nvidia Accelerated Computing, we study the noise
Impact of ever larger quantum chip designs.”
Understanding noise in quantum hardware designs requires complex dynamic simulations able to fully capturing how qubits in a quantum processor interact with their environment.
Running these simulations has traditionally been prohibitively computationally intensive. However, using the CUDA-Q platform, Google can deploy 1,024 Nvidia H100 Tensor Core GPUs within the Nvidia Eos supercomputer to perform one in all the world's largest and fastest dynamic simulations of quantum devices – at a fraction of the associated fee.
“The power of AI supercomputers will contribute to the success of quantum computing,” said Tim Costa, director of quantum and HPC at Nvidia, in a press release. “Google’s use of the CUDA-Q platform demonstrates the central role of GPU-accelerated simulations in advancing quantum computing to resolve real-world problems.”
With CUDA-Q and H100 GPUs, Google can perform fully immersive, realistic simulations of 40-qubit devices – the most important simulations of its kind performed. Thanks to the simulation techniques provided by CUDA-Q, noisy simulations that will have taken per week can now be carried out in be carried out in a number of minutes.
The software that powers these accelerated dynamic simulations will probably be publicly available on the CUDA-Q platform, allowing quantum hardware engineers to quickly scale their system designs.
