Quantum computing hardware and software company IonQ says it has developed a quantum error-correction decoder that runs in real time on a single conventional CPU, potentially reducing the classical computing overhead that can slow larger quantum systems. In simulations spanning up to 408 logical qubits, the company’s decoder added as little as 0.02% processing time under standard operating noise conditions.
Quantum computing gains its power through qubits, a special type of ‘bit’ that can have any of many states between the traditional binary on/off state of classical computing. The primary problem with qubits is that these ‘superimposed’ conditions that are neither on nor off are extremely fragile to environmental noise and very easily collapse, reverting to the binary state useless to quantum computing.
We call operational qubits ‘logical’ and the much larger number of additional qubits needed to find a collapsed ‘logical’ qubit as ‘physical’ qubits. The current solution is to have many more physical qubits able to locate each faulty (collapsed) logical qubit – but it requires slower classical computing to decode the information provided through the larger number physical qubits.
This process adds delay to the speed and power provided by the quantum computer, but the successful end result is more fault tolerant and therefore more powerful quantum computing. But it introduces an additional ironic tradeoff in continuing to build more powerful quantum computers. More powerful quantum computing is generally the result using more logical qubits; but each new logical qubit requires a large number of extra physical qubits to maintain error correction.
The growth in the number of qubits together with their fragility will inevitably require an increase in the delay caused by classical quantum error decoding.
IonQ has announced a new solution to what it describes as this major bottleneck introduced by the classical computer– a new end-to-end real-time quantum error correction decoder running on a single standard classical computing CPU. In its own research published on arXiv, IonQ simulated up to 408 logical qubits across 88 memory blocks and magic factories (functional zones separately handling data storage and complex calculations).