英伟达在德国于利希JUPITER超算完成104量子比特容错量子计算模拟：规模创纪录，物理量子比特需求减少90%
英伟达（NVIDIA）10月19日宣布，其研究人员在位于德国于利希超级计算中心的欧洲首台百亿亿次级超级计算机JUPITER上，完成了一次创纪录的容错量子计算模拟：104个量子比特的模拟是迄今对容错量子处理器规模最大、最完整的模拟，验证了名为“魔态培养”（magic state cultivation）的前沿容错方案，该方案所需物理量子比特比以往方法减少90%。
模拟在搭载英伟达GH200超级芯片的系统上运行，规模扩展至8,192块GPU。此前于利希团队曾在理想无噪声条件下模拟50个量子比特，而此次模拟在真实硬件噪声条件下运行，帮助研究人员理解误差对容错计算的影响，从而在制造硬件之前先优化设计。英伟达称，这类模拟把通往实用量子计算机的路径缩短了数年，应用前景包括药物发现、金融建模与材料科学。
英伟达创始人兼首席执行官黄仁勋表示：“有用的量子计算需要极其复杂的容错系统——理解它们在大规模下的行为，是建造它们的前提。我们在于利希取得的成果表明，量子处理器与GPU超级计算机可以协同工作，攻克原本遥不可及的研究问题。”于利希超级计算中心主任克丽丝特尔·米希尔森（Kristel Michielsen）表示，欧洲对JUPITER的投资让研究人员得以探索以往无法触及的问题。
该项目先在英伟达加速量子研究中心（NVAQC）的GB200 NVL72系统上开发，将测试从4块GPU扩展到数百块GPU，把原需数周的模拟缩短到数天；模拟软件采用英伟达cuQuantum库与CUDA-Q平台的容错逻辑层。
NVIDIA Runs Record 104-Qubit Fault-Tolerant Quantum Simulation on Europe's JUPITER Supercomputer, Cutting Qubit Needs by 90%
NVIDIA announced on October 19 that its researchers have run a record-breaking fault-tolerant quantum computing simulation on JUPITER, Europe's first exascale supercomputer at the Jülich Supercomputing Centre in Germany. The 104-qubit simulation is the largest comprehensive simulation of a fault-tolerant quantum processor to date, and it validates magic state cultivation, a leading fault-tolerant approach that requires 90% fewer physical qubits than previous methods.
The simulation ran on NVIDIA GH200 superchips and scaled to 8,192 GPUs. Earlier Jülich research had simulated 50 qubits under ideal, noiseless conditions; this work models magic-state cultivation under realistic hardware noise, giving researchers data to predict the performance of future designs before hardware is built. NVIDIA says the approach can shorten the path to useful quantum computers by years, with applications in drug discovery, financial modeling and materials science.
"Useful quantum computing will require fantastically complex, fault-tolerant systems — and understanding how they behave at scale is essential to building them," said Jensen Huang, founder and CEO of NVIDIA. "Our results from Jülich show how quantum processors and GPU supercomputers can work together to tackle research that would otherwise be out of reach." Kristel Michielsen, director of the Jülich Supercomputing Centre, said Europe's investment in JUPITER lets researchers investigate questions previously beyond reach.
The project was developed at NVIDIA's Accelerated Quantum Research Center (NVAQC) on a GB200 NVL72 system, scaling tests from four GPUs to hundreds and cutting a weekslong simulation effort to days. The simulations used the NVIDIA cuQuantum software libraries and the CUDA-Q platform's fault-tolerant logical layer.
来源：GSM·美国（原文链接）
图片来源：超级计算机数据中心（资料图）。Photo: Ana Las Heras, CC BY-SA 4.0, via Wikimedia Commons
