NVIDIA Releases cuQuantum SDK v25.11, Boosting Error Correction and Large-Scale Quantum Simulation
NVIDIA has released the latest version of the cuQuantum SDK, v25.11, a collection of high-performance libraries designed to accelerate GPU-based quantum computing simulations. Pauli propagation and stabilizer simulations, which are essential for simulating massive quantum computers, quantum error correction, and producing AI training data, are two new significant workloads that this upgrade adds components for. According to the study, experimental quantities that are otherwise too complicated for precise classical simulation, like expectation values, can be efficiently estimated with the help of the cuQuantum cuPauliProp package. Additionally, a tool to greatly increase the throughput and sampling rates for frame simulations—which are crucial for evaluating quantum error correcting codes—is presented: the cuQuantum cuStabilizer library.
Pauli propagation (cuPauliProp) and stabilizer simulations (cuStabilizer) are two new high-performance libraries designed to speed up the modelling of large-scale quantum computers. NVIDIA has announced the release of version 25.11 of its cuQuantum SDK. These novel elements are thought to be essential for simulating and verifying large-scale quantum systems.
You can also read NVIDIA CUDA-X libraries power quantum with QuTiP-cuQuantum
The Validation Challenge
Validating the results becomes crucial to guaranteeing the outputs can be relied upon as the quality of quantum processing units (QPUs) keeps improving and devices scale beyond what is conventionally simulable. Two main workload classes that are essential for algorithm engineering, verification, and validation for intermediate to large-scale quantum devices are covered by the new libraries in cuQuantum SDK v25.11.
Simulating Intractable Observables with CuPauliProp
In order to effectively replicate the observables of large-scale quantum circuits, especially those that contain noise models of actual quantum processors, Pauli propagation is a relatively novel technique.
- Managing Intractability: The computation of expectation values is essential to many important quantum applications, including the Variational Quantum Eigensolver (VQE) and the quantum simulation of physical dynamics. It can be unaffordable to simulate these expectation values precisely using traditional methods. By representing states and observables as weighted sums of Pauli tensor products and dynamically eliminating terms that significantly affect the intended expectation value, Pauli propagation enables estimate.
- Particular Uses: The method has shown promise in the simulation of near-Clifford or noisy circuits. Additionally, it has shown remarkable performance when modelling circuits that simulate the evolution of specific quantum spin systems, such as the “utility circuits” referenced in relation to IBM’s 127-qubit tests.
- Performance: By speeding up Pauli propagation on NVIDIA GPUs, developers can enhance classical circuit simulation with new tools. Speedups of several orders of magnitude have been demonstrated in tests using NVIDIA DGX B200 GPUs in comparison to single-threaded Qiskit Pauli-Prop running on modern data center CPUs.
You can also read Qiskit SDK v2.2 Introduces Powerful Qiskit C API for Quantum
CuStabilizer: Error Correction Testing at High Throughput
The Gottesman-Knill theorem, which states that circuits made entirely of Clifford gates (CNOT, Hadamard, and Phase gates) may be effectively simulated classically in polynomial time, is the foundation of stabilizer simulations. This capacity is essential for the large-scale testing of quantum error-correcting (QEC) codes and resource estimation.
- Acceleration of Frame Simulation: The cuStabilizer library concentrates on increasing a frame simulator’s throughput for sampling speeds. Frame simulation simulates the impact of quantum noise on the quantum state since quantum devices are not flawless. The frame simulator monitors the change brought about by adding arbitrary noisy gates if the noise-free outcome is known.
- Meeting Data Requirements: Because the number of possible noisy gate combinations rises quickly with circuit size, accurately simulating error correction algorithms necessitates a large number of “shots” or samples. For users creating QEC codes, testing new decoders, or creating the large datasets required for AI decoder training, cuStabilizer is therefore perfect.
- Performance: Compared to cutting-edge CPU codes like Google Stim, users can significantly increase sampling throughput by utilising GPU acceleration. With demonstrated speedups of up to 1,060x for simulations using large code distances (e.g., surface code distance 31) on the NVIDIA DGX B200 GPU, optimal performance is attained for large quantities of samples and qubits.
Users can install cuPauliProp and cuStabilizer using pip.
You can also read NVIDIA NVQLink To Connect Quantum Processors With GPUs
