This collaboration advances real-time quantum control and demonstrates the integration capabilities of Diraq's silicon qubit arrays.
Diraq and QM Technologies Inc. announced the first-ever tight integration of GPUs and silicon quantum processors with the deployment of a system using the NVIDIA DGX Quantum architecture to achieve real-time communication speeds of 3.3 microseconds between quantum hardware and NVIDIA Grace-Hopper superchips, orchestrated through QM’s OPX1000 hybrid controller.
Working at Diraq’s Sydney laboratories, engineers from Diraq and QM demonstrated how GPU acceleration can solve critical bottlenecks in quantum computing. The team implemented three applications that address fundamental scaling challenges: a real-time readout enhancement that was previously only possible through post-processing, automated calibration using machine learning, and accelerated quantum state initialization.
The results were achieved within just one week of DGX Quantum installation at Diraq’s facilities, demonstrating the system’s ability to have an immediate impact on quantum computing research. The system combines NVIDIA Grace Hopper with QM’s OPX1000 control system, enabling quantum processors to leverage AI and other GPU-accelerated computation in real-time.
The most significant achievement was implementing correlated readout in real-time, building on methods from Diraq’s published research. This complex signal processing task requires computational power beyond what traditional FPGAs can provide, making GPU acceleration essential. Machine learning algorithms running on the GPU also automated calibration processes that traditionally require hours of manual tuning, while Diraq’s algorithmic initialization protocols (detailed in their recent Nature publication) used GPU acceleration to achieve faster state preparation, which is critical for ensuring quantum states can be prepared before they degrade.
The 3.3-microsecond round-trip time between the quantum processor and GPU is crucial for these achievements. This low latency enables real-time feedback loops, sending commands and receiving responses before quantum information degrades, which is essential for practical quantum computing.
Looking ahead, DGX Quantum will be able to support real-time error-correction protocols on a hybrid quantum–classical architecture. DGX Quantum also enables Diraq’s hardware to be directly integrated with NVIDIA’s open-source quantum development platform, NVIDIA CUDA-Q, highlighting the viability of GPU-based control for future quantum computers.
The results will be highlighted at GTC Paris 2025, where NVIDIA’s Director of Quantum Algorithm Engineering, Elica Kyoseva, will discuss the future of hybrid quantum-classical algorithm development.
For more information, visit diraq.com.
