Honeywell's quantum scientists have quadrupled the capabilities of the company's quantum computer, with the device achieving record levels of performance less than a year after the first generation of the system was released.
The System Model H1, a ten-qubit quantum computer, effectively reached a quantum volume of 512 – four times as much as was attained in the previous tweak of the system, which saw the H1 reach a quantum volume of 128.
Released commercially last June (at the time as the System Model H0), the H1 makes use of trapped ions, unlike IBM and Google's devices, which are built with superconducting qubits. Honeywell's new record is eight times as much as was achieved with the System Model H0, which launched with a quantum volume of 64.
Quantum volume is a concept that IBM developed in 2017 as a way of measuring various aspects of a quantum computer's performance; in simple terms, the higher the quantum volume, the higher the potential for resolving real-world problems across industry and research. Designed to be independent of the architecture of any given quantum computer, quantum volume can measure any system that runs quantum circuits.
SEE: Hiring Kit: Computer Hardware Engineer (TechRepublic Premium)
For example, one measurement that is indicative of a quantum computer's capabilities is qubit fidelity, which is critical to understanding how well a device can implement quantum code. According to Honeywell, the average single-qubit gate fidelity in the latest version of the H1 was 99.991%.
The final number that determines quantum volume is an aggregate of many other measurements and tests of a single quantum system's operations: they include the number of physical qubits in the quantum computer, but also the device's error rate, and connectivity, which reflects the extent to which qubits can be fully connected to each other within the device.
This is why it is possible for a quantum system to reach a high quantum volume, even with few qubits. Despite having only ten qubits, for instance, Honeywell's System Model H1 performs well when it comes to error rates and connectivity, which has earned the device a top spot for its overall capabilities. In comparison, last year IBM's 27-qubit client-deployed system achieved a quantum volume of 64.
The new milestone, therefore, has prompted Honeywell's president of quantum solutions Tony Uttley to describe the System Model H1 as "the highest performing quantum computing system in the world."
Honeywell has made no secret of its strategy, which consists of focusing on qubit fidelity and connectedness, before attempting to scale up the number of qubits. "When you hear about fidelity and error, that's about the quality of the quantum operation," Uttley told ZDNet. "It's about knowing how often you get the right answer when you run these quantum algorithms."
"We have taken one approach that is very unique when it comes to how to get the most out of these near-term systems," he continued. "Nobody is talking about millions of qubits right now – we're talking about tens of qubits. To get the most out of these tens of qubits, you have to have super-high fidelity, fully-connected and highly-controlled systems. That's our approach."
Making these highly reliable systems available to Honeywell's customers now enables businesses to test and trial with small-scale applications while waiting for the company to design and build new generations of more capable quantum computers, according to Uttley.
Honeywell recently introduced the first subscription-based plan for the usage of the H1, which grants paying customers a monthly access to the machine.
With only ten qubits, there is little that the device can achieve on top of proofs of concepts, designed to be implemented in full scale once a larger computer is available; but high-profile customers are nevertheless flocking to Honeywell's services.
J.P. Morgan Chase, for example, is investigating how the company's quantum computer might improve operations in banking; and BMW is piloting the use of Honeywell's hardware to optimize supply chains for car manufacturing.