NVIDIA and Google's Quantum Leap: Revolutionizing Quantum Computing with Eos Supercomputer
1min read In a significant stride into the realm of quantum computing, NVIDIA (NVDA.US), the chip giant at the core of the global generative artificial intelligence surge, has announced a strategic collaboration with Google's Quantum AI division. This partnership aims to address some of the most complex challenges in quantum computing, a field often deemed "science fiction" due to its advanced nature. NVIDIA's high-performance computing capability in AI GPUs, coupled with its expertise in parallel architecture design and accelerated software ecosystems, positions it uniquely in this venture.
NVIDIA's state-of-the-art "Eos Supercomputer" is set to play a pivotal role in developing quantum chips by combining its advanced computational power with Google's research prowess. The Eos Supercomputer is tailored for quantum chip simulation, peripheral control hardware design, algorithm verification, and the enhancement of software toolchains, providing a robust environment for quantum development.
On November 19, both companies declared their alliance, emphasizing the role of NVIDIA's CUDA-Q platform in aiding Google's Quantum AI researchers to develop extensive digital models of quantum computers. These models are crucial in overcoming design challenges inherent in quantum technology.
Through this collaboration, Google's Quantum AI unit plans to use NVIDIA's Eos Supercomputer and hybrid quantum-classical computing platforms to simulate the physical properties of its quantum processors. This approach is expected to help Google mitigate the limitations of quantum computing hardware, such as "noise," which refers to the requirement for quantum computations to halt after a certain number of operations.
Guifre Vidal, a research scientist at Google's Quantum AI team, highlighted the importance of controlling noise while scaling quantum hardware, noting that this control is vital to developing commercially viable quantum computers. Understanding noise in quantum hardware design necessitates complex dynamic simulations capable of fully capturing interactions between qubits and their environments.
To facilitate these simulations, NVIDIA has provided a supercomputer powered by 1,024 H100 Tensor Core GPUs, designed to execute one of the world's largest and fastest dynamic simulations of quantum devices. This setup significantly reduces costs while vastly improving simulation speed and accuracy.
With the CUDA-Q and H100 GPU resources at their disposal, Google can perform detailed and realistic dynamic simulations of devices containing 40 qubits. These simulations, previously taking up to a week, can now yield results in mere minutes, marking a groundbreaking enhancement in both time efficiency and cost reduction.