Quantum Machines connects its controller to NVIDIA’s error-correction software
A test sends simulated error checks out for interpretation and returns a correction decision during the same run.
Reading tools for text size, copying, printing and your reading list

has a timing problem: identifying a fault is useful only if the answer arrives when the computation needs it. Quantum Machines has tested a connection that sends error-check data from its controller to software that interprets the pattern and returns a correction decision.[1]
On November 16, the company reported connecting its OPX1000 controller to NVIDIA’s CUDA-Q QEC decoding server. A decoder interprets error-check patterns to infer the needed correction. This test replayed simulated data rather than freshly measured errors. PyMatching ran on the CPU of an NVIDIA GH200 system; GPU dispatch remains future work.[1]
The company reports 2.4 microseconds for two-way communication. Decoding time was measured separately and increased with code distance, a parameter governing the error-correcting code’s size and protection. The communication figure is not a measured complete correction time.[1]
Each tested configuration used 10,000 repeated runs and simulated error-check records generated with an assumed 0.1% physical error probability. The correction decision returned within the same run. Early installation for OPX1000 customers is planned for December 2026.[1]
The software arrangement matters beyond a single stopwatch result. NVIDIA’s server guide separates the program controlling a processor from the service predicting corrections. A compatible controller sends measurements and receives logical correction decisions. Its configuration identifies the decoder and how to interpret the measurements, letting researchers change the decoding pipeline without treating every experiment as an entirely separate system.[2]
The guide also makes a crucial distinction: acknowledging receipt of measurements does not mean decoding has finished. Its examples test functionality, not a guaranteed quantum-processor feedback time. That is why the useful next result is a complete, clearly bounded test using measured qubit errors, with response-time variation and correction quality reported together. Faster communication alone cannot establish better protection of quantum information.[2]
Keep your reading trail
Keep up with these topics
Explore the subjects covered in this story.