Benchmarking Zero-Setup Quantum Circuit Simulators
2026-07-10 • Distributed, Parallel, and Cluster Computing
Distributed, Parallel, and Cluster ComputingPerformance
AI summaryⓘ
The authors studied how well different tools simulate quantum systems using GPUs compared to CPUs. They tested two common methods—matrix product states (MPS) and Pauli path simulation (PPS)—on various platforms including BlueQubit and AWS Braket. They found that GPUs often run simulations much faster, especially for large quantum problems, and some GPU-based systems can reach higher accuracy levels that CPU-based ones cannot. The authors also shared all their testing code for others to use and verify.
GPU accelerationquantum simulationmatrix product states (MPS)Pauli path simulation (PPS)benchmarkingquantum computingAWS Braketaccuracy thresholdsbond dimensionIBM kicked Ising model
Authors
Arul Rhik Mazumder, Mohammed Zuhair Mullath, Hayk Tepanyan
Abstract
Practitioners increasingly rely on hosted simulation environments, but their performance characteristics remain poorly documented. We present a systematic benchmarking study of GPU-accelerated approximate quantum simulation across two widely used methods: matrix product states (MPS) and Pauli path simulation (PPS), comparing BlueQubit (a hosted tool that handles hardware provisioning, simulator configuration, and job orchestration) against AWS Braket, Quantum Rings, PPS-Qiskit, and PauliPropagation.jl. For MPS, we find that GPU runtime yields sub-quadratic scaling with bond dimension, with a growing advantage over CPU at increasing scale. For Pauli path simulation on IBM's 127-qubit kicked Ising benchmark, GPUs deliver up to $1{,}400\times$ speedup at fine truncation thresholds ($δ= 2.5 \times 10^{-5}$, 27.6M Pauli terms), and are the only backends that reach accuracy regimes below $δ= 10^{-5}$, which remained inaccessible to the commodity CPU-based implementations and self-contained SDKs evaluated here. We also provide a reproducible characterization of these simulators across regimes, including tradeoffs that isolated evaluations do not show. To support transparency and reuse, we provide a public GitHub repository containing all benchmarking code and configurations.