Quantum · Article
Quantum Machines Shows Sub-Microsecond Quantum Error Decoding on OPX1000 With NVIDIA
Quantum Computing Report ·

At a glance
Quantum Machines showed that its OPX1000 quantum control system can decode quantum error corrections in the middle of a quantum circuit. It used NVIDIA's CUDA-Q QEC decoding server and NVQLink on a GH200 server, with decoding under one microsecond per round for code distances 7 and 11. Early customer access is planned for December 2026.
Key points
- The test used pre-sampled surface-code error data at a 0.1% physical error rate on an NVIDIA GH200 Grace Hopper server.
- The round-trip time between the controller and the decoder was 2.4 microseconds.
- Decoding took under one microsecond per round for code distances 7 and 11.
- Decoders such as PyMatching and NVIDIA's fusion decoder can be swapped through a YAML configuration edit.
Why it matters
Fixing quantum errors during a running circuit is a key step toward useful quantum computers, so this speed result matters for the field.
Original source
Quantum Machines Demonstrates Sub-Microsecond Surface-Code Decoding on OPX1000 Using NVIDIA CUDA-Q QEC and NVQLink
Quantum Computing Report · Nov 18, 04:03 KST
Read on Quantum Computing Report ↗AI summarized the original article. Check the original for accuracy. Copyright belongs to Quantum Computing Report.
More from this field
- ICFO Quantum Microscope Images All 10 Spin States of Strontium Atoms in One Shot
Researchers at ICFO in Spain built a quantum-gas microscope that can image strontium-87 atoms and tell apart all 10 of their spin states in a single run. Each spin state shows up as a different color. The team reported the result in Physical Review X.
Phys.org · Nov 18, 09:00 KST - Nvidia Teams Up with IBM and Fermilab on Quantum Error Decoding, Claims 13,000x Speedup
Nvidia said it will work with IBM and the Fermi National Accelerator Laboratory on quantum computing research. Nvidia also released a new decoding algorithm called TCBP-JOBS that it says is up to 13,000 times faster at correcting quantum errors.
AI 매터스 · Nov 18, 08:17 KST - QUDORA Adds NVIDIA CUDA-Q to Qamelion Emulator and Starts German-Funded QUANDA Project
QUDORA Technologies, a trapped-ion quantum company in Braunschweig, Germany, added NVIDIA CUDA-Q support to its Qamelion emulator, which lets developers test quantum-classical programs before running them on real hardware. The company also started a three-year German government-funded project called QUANDA to make its trapped-ion systems easy to use inside supercomputing centers.
Quantum Computing Report · Nov 18, 04:29 KST - Quantum Observers Can Disagree on Outcomes, New Theorem Shows Limits
Two observers can share the same quantum description and rules yet still disagree about an outcome's chance, even while trusting each other's estimates. A new study by researchers in Spain and the US shows that this disagreement can persist because a measurement can change the system itself. Shared classical records of measurement results can restore agreement.
Phys.org · Nov 18, 04:20 KST