The findings show that quantum computing hardware has improved over time and that application-specific benchmarks can serve as a more practical yardstick for comparing the capabilities of alternative types of quantum hardware.
“This is clear evidence that things are heading in the right direction for quantum optimisation,” said Jack Baker, a quantum algorithms researcher at Aqnostiq. “With quantum hardware receiving increasing interest and investment every year, these performance increases are poised to accelerate.”
With the steady increase of quantum bits (qubits) made available by hardware providers, the relative performance and practical value of quantum computers has been difficult to assess. General benchmark studies have been conducted and deemed inconclusive, as they are not predictive of performance.
Agnostiq employed application-specific benchmarks and conducted its research using a portfolio optimisation task to determine whether quantum computers have actually improved over time for specific use cases.
As a consequence, the team discovered that the performance of gate-model quantum computers has improved for performing optimisation problems, which is further indication the field is on a path towards commercialisation. For more complex variations of the algorithm, the team also discovered that solution quality can improve as more quantum resources are added. This was a previously unseen result.
According to the team at Agnostiq, these findings should encourage organisations who depend on large scale optimisation, simulation, or machine learning for mission critical tasks to invest in quantum computing technologies.
- High quality portfolios were produced using quantum circuits requiring larger numbers of gates (operations on the qubits) than previously demonstrated. Since increasing the number of gates produces more noise, this shows the quality of hardware has improved for performing combinatorial optimization.
- The peak solution quality was observed at higher depth (p=4) on 3 qubits on an IonQ trapped ion machine.
- As a non-trivial effect of studying application dependent performance, an IBM machine with the lowest qubit quality (quantum volume = 8) performed best of all the IBM machines tested.
- Quantum computers presently give variable results depending on the time they were accessed. Variability needs to be considered with all benchmarking numbers, as it can be as high as 29 percent.
“We are at an interesting point where every hardware paradigm has its own set of performance metrics that they are optimising against, and each of them is improving across different dimensions” said Agnostiq’s head of R&D, Santosh Kumar Radha. “We recognised a need to better understand how these non-trivial improvements translate to real-world applications.”
The global market for quantum computing is expected to reach $5 billion by 2028 but while quantum computing has the capabilities to speed up computations it remains largely inaccessible to organisations, due mainly to the novelty of the technology and the high level of expertise required to build applications.
Agnostiq is building a suite of tools to lower the barriers for enterprises to enter into the world of quantum computing.
Recently released, Covalent is the first open-source workflow orchestration platform designed specifically for quantum and high-performance computing. It aims to make quantum and high-performance computing resources more accessible to practitioners, including researchers, machine learning engineers and data scientists.
The company has secured $5 million in funding to date. Investors include Differential Ventures, Scout Ventures, Boost VC, Tensility Venture Partners, and Green Egg Ventures.