Oracle Cloud Infrastructure has tested a financial option-pricing circuit ranging from 28 to 32 qubits using Classiq’s quantum software platform, NVIDIA CUDA-Q and an NVIDIA A100 80GB Tensor Core GPU.
The benchmark focuses on statevector simulation, where GPU memory capacity and bandwidth are important factors. The tested workload estimates the price of a European call option using Iterative Quantum Amplitude Estimation, combining quantum state preparation, an oracle encoding the payoff condition and Grover-style amplitude amplification.
28 to 32 Qubits: Execution Time Scales with Statevector Size
Classiq generated the quantum circuit from a higher-level functional description and emitted OpenQASM 3.0. NVIDIA CUDA-Q then compiled and executed the resulting quantum kernel using the cuStateVec backend.
Because the state space doubles with every additional qubit, increasing the circuit from 28 to 32 qubits increases the statevector size by a factor of 16. A 28-qubit statevector contains approximately 268 million complex amplitudes, while a 32-qubit statevector contains approximately 4.3 billion.
At 28 qubits, execution completed in 2.2 seconds. At 32 qubits, it completed in 22.1 seconds. In the tested configuration, this was an approximately tenfold increase in execution time for a sixteenfold increase in statevector size.
GPU Simulation Shows a Significant Speed Advantage
The tested GPU configuration was approximately 120 times faster than the specified CPU configuration at 28 qubits and 77 times faster at 29 qubits.
These figures are specific to the hardware, software, precision and benchmark settings used in the test and should not be generalised to other CPU or GPU systems.
The results also highlight the importance of memory characteristics in statevector simulation. At 32 qubits, the workload remained within the memory capacity of a single NVIDIA A100 80GB GPU. At 33 qubits, the raw single-precision statevector would require approximately 64 GiB, leaving less capacity for simulator workspace and other overhead.
Larger simulations may require distributed multi-GPU execution or different numerical representations, with potential effects on performance and numerical precision.
Classiq, CUDA-Q and OCI Form the Simulation Workflow
The workflow consisted of three main layers. Classiq generated and optimised the circuit, NVIDIA CUDA-Q compiled the OpenQASM representation into a quantum kernel and executed it on the GPU, while OCI provided the underlying compute infrastructure.
The benchmark also showed the value of reusing previously synthesised circuits. At 28 qubits, a run including synthesis and compilation took 6.2 seconds, compared with 2.2 seconds for execution of a previously synthesised circuit.
For repeated development and testing, avoiding unnecessary synthesis and compilation can therefore reduce end-to-end latency.
A Benchmark for Quantum Algorithm Development
The results do not represent fault-tolerant quantum hardware or production quantum advantage. The 32-qubit noiseless simulation is a development tool for teams writing and validating quantum algorithms before large-scale physical quantum hardware becomes available.
The option-pricing circuit is also a relatively well-behaved benchmark, so the results should not be generalised to all quantum simulation workloads. Different circuit structures and memory access patterns can produce different scaling behaviour.
Oracle’s next planned extension is a 36-qubit QAOA portfolio optimisation workload running on a DGX A100 node on OCI. The longer-term work will examine how statevector simulation scaling changes on newer GPU platforms, including Blackwell and Vera Rubin systems.
For teams developing quantum applications today, the tested 28 to 32 qubit range demonstrates that noiseless simulation on a single A100 can support an iterative development workflow, with execution times of a few seconds at 28 qubits and around 22 seconds at 32 qubits.
Source: Oracle Cloud Infrastructure
