Chinese researchers have achieved a significant milestone in quantum computing by successfully simulating Google’s 53-qubit Sycamore quantum circuit using 1,432 NVIDIA A100 GPUs and innovative algorithmic approaches. This achievement represents a direct challenge to Google’s 2019 claim of quantum advantage, where their Sycamore processor was said to perform calculations that would take traditional supercomputers thousands of years to complete.
The research team developed sophisticated tensor network contraction techniques to efficiently estimate quantum circuit output probabilities. They implemented a slicing strategy that breaks complex tensor networks into smaller, manageable parts, significantly reducing memory requirements while maintaining computational efficiency. Additionally, they employed a “top-k” sampling method that focuses on the most probable bitstrings, improving the linear cross-entropy benchmark (XEB) while reducing computational demands.
To validate their approach, the researchers conducted experiments with smaller-scale random circuits, demonstrating excellent agreement between their results and theoretical predictions. They further optimized performance by refining tensor index ordering and minimizing communication between GPUs.
The most striking result is that their classical simulation generated three million uncorrelated samples with higher XEB values in just 86.4 seconds, approximately seven times faster than Google’s Sycamore processor which required 600 seconds for the same task. Moreover, the classical approach consumed only 13.7 kWh of electricity, compared to Sycamore’s 4.3 kWh for cooling alone.
This breakthrough provides the first unambiguous experimental evidence refuting Sycamore’s quantum advantage claim. It also redefines the boundary between classical and quantum computational capabilities in the realm of random circuit sampling. The research demonstrates that with algorithmic innovations and optimized hardware utilization, classical systems can still compete with early quantum computers in specific tasks.
As this field evolves, we can expect continued competition between improved classical algorithms and more advanced quantum hardware, progressively establishing clearer benchmarks for genuine quantum computational advantage.
Reference: “Leapfrogging Sycamore: harnessing 1432 GPUs for 7× faster quantum random circuit sampling” by Xian-He Zhao, Han-Sen Zhong, Feng Pan, Zi-Han Chen, Rong Fu, Zhongling Su, Xiaotong Xie, Chaoxing Zhao, Pan Zhang, Wanli Ouyang, Chao-Yang Lu, Jian-Wei Pan and Ming-Cheng Chen, 12 September 2024, National Science Review. DOI: 10.1093/nsr/nwae317
