The challenge of simulating complex physical systems drives innovation in both quantum and classical computing, and researchers are now directly comparing the performance of both approaches. Philipp Hanussek from Sorbonne Université, Jakub Pawłowski from Wrocław University of Science and Technology, and Zakaria Mzaouali from Universität Tübingen, along with Bartłomiej Gardas and colleagues, present a new benchmarking suite inspired by real-world physical dynamics to test the capabilities of quantum and classical annealers rigorously. The team converts the problem of simulating how systems evolve into a format suitable for these machines, allowing for a direct comparison of their speed and accuracy. They stress-test the process using a diverse set of models. This work demonstrates significant advances in analogue optimisation, particularly with the latest generation of quantum annealers, while also highlighting the continued strength of classical algorithms, and establishes a valuable framework for tracking future progress in this rapidly evolving field.
Time-to-Solution on Quantum and Classical Hardware
This supplemental material details a study comparing the performance of different computing platforms — quantum annealers and classical computers — when solving quantum-inspired optimisation problems. The core focus is on time-to-solution (TTS99), the time required to find a solution with 99% probability, a crucial metric for practical application. The research evaluates several hardware options, including the VeloxQ and Advantage quantum annealers, as well as standard and graphics processing units running simulated annealing algorithms. Key findings reveal that the time-to-solution scales exponentially with the problem size, a common characteristic of optimisation challenges.
The exponent of this scaling, denoted as β, is critical; a larger β indicates better performance on larger problems. Results demonstrate that graphics processing units significantly outperform standard CPUs for simulated annealing, particularly as problem size increases, due to the parallel processing capabilities of GPUs. The Advantage2 1. 6 quantum annealer shows promising results, suggesting the potential of quantum annealing for these types of problems, although it sometimes fails to find a solution within a reasonable timeframe. The study utilises eight different problem instances, varying in complexity, to assess the performance of each platform comprehensively.
Benchmarking Quantum and Classical Dynamics with QUBO
Researchers have developed a new benchmarking suite inspired by physical dynamics to rigorously compare quantum and classical computers. They converted the real-time evolution of a quantum system into a quadratic-unconstrained binary optimisation (QUBO) problem, allowing both quantum annealers, specifically D-Wave systems, and classical solvers to tackle the same computational challenge. This approach uses eight representative models, encompassing single-qubit rotations, multi-qubit entangling gates, and non-Hermitian generators, to comprehensively test the workflow. The method employs a parallel-in-time encoding, discretising time evolution into sequential slices and assembling them into a clock Hamiltonian.
Researchers submitted these QUBO instances to D-Wave Advantage and Advantage2 quantum annealers, as well as the classical solver VeloxQ and simulated annealing. This direct comparison eliminates confounding factors, allowing for a clear assessment of scaling behaviour. Experiments tracked both success probability and time-to-solution. Results demonstrate that D-Wave Advantage2 consistently outperforms its predecessor, exhibiting an order of magnitude higher ground-state success probability and a smaller time-to-solution exponent. However, VeloxQ currently maintains the overall lead, highlighting the maturity and strength of optimised classical algorithms. This parallel-in-time QUBO framework establishes a versatile testbed for tracking progress toward competitive dynamics simulation, enabling researchers to evaluate the potential for quantum advantage rigorously.
Quantum Benchmarking via Dynamical System Simulation
Researchers have developed a new benchmarking suite inspired by physical dynamics to compare the performance of quantum and classical computers rigorously. The team converted the real-time evolution of quantum systems into a quadratic-unconstrained binary optimisation (QUBO) problem, enabling execution on quantum annealers and classical solvers. This innovative approach allows for a direct, like-for-like comparison, removing confounding factors that previously hindered accurate performance assessments. Experiments involved eight representative quantum models, encompassing single-qubit rotations, multi-qubit entangling gates, and non-Hermitian generators, to comprehensively stress-test the workflow.
Results demonstrate that the D-Wave Advantage2 quantum annealer consistently surpasses its predecessor in terms of both ground-state success probability and time-to-solution, indicating significant hardware advancements. Specifically, Advantage2 achieves an order-of-magnitude improvement in ground-state success probability compared to the earlier generation. However, the state-of-the-art classical solver, VeloxQ, currently retains the overall lead on the problem sizes accessible today, highlighting the continued strength of optimised classical algorithms. The research establishes a parallel-in-time QUBO framework as a versatile testbed, providing a clear trajectory for tracking and evaluating progress toward competitive quantum dynamics simulation. This framework enables researchers to map continuous-time quantum dynamics to a solver-agnostic combinatorial problem, allowing for unbiased performance comparisons between quantum and classical optimisation hardware.
D-Wave and VeloxQ Benchmarked via Quantum Dynamics
This research presents a new method for benchmarking quantum annealers and classical solvers by translating the problem of continuous quantum dynamics into a series of binary optimisation tasks. The team successfully demonstrated this approach using the D-Wave Advantage2 processor and a classical solver called VeloxQ, allowing for a direct comparison of their performance across a diverse set of dynamical problems. Results indicate that the Advantage2 processor represents a significant improvement over its predecessor, achieving a higher success rate across various annealing schedules. While the D-Wave system showed marked progress, VeloxQ currently maintains the fastest overall runtime, highlighting the continued strength of classical optimisation techniques.
The study acknowledges that the performance balance between hardware advancements and algorithmic improvements is dynamic, with leadership potentially shifting between the two approaches. The authors note that their encoding method scales favourably, requiring additional qubits proportional to the number of simulated time steps, suggesting the potential for simulating more complex systems as quantum hardware improves. The researchers have released an open-source benchmark suite to facilitate ongoing evaluation of both quantum and classical optimisation methods. They recognise that current quantum devices still require further development to consistently outperform the best classical heuristics, but anticipate that denser qubit lattices and wider connectivity will enable medium-scale dynamical simulations in the future.