Qblox measures microsecond round trips between quantum control and GPUs

Qblox announced measured GPU-control integration on October 22. The company reports round-trip latency below 4 microseconds with an NVIDIA RTX PRO 6000 Blackwell GPU and below 7.2 microseconds with DGX Spark, using NVQLink.

Fast classical feedback is important for calibration and error correction while a quantum experiment is still running. Moving data to a GPU only helps if the answer returns within the timing budget of the quantum hardware.

These are measured integration latencies, not a direct measure of a logical qubit's fidelity. Keeping the two hardware configurations separate also avoids assigning the faster workstation result to the smaller DGX Spark system.