Commercial quantum computer leaves PC in the dust

[frigidmagi]

NewScientist

For the first time, a commercially available quantum computer has been pitted against an ordinary PC – and the quantum device left the regular machine in the dust.

D-Wave, a company based in Burnaby, Canada, has been selling quantum computers since 2011, although critics expressed doubt that their chips were actually harnessing the spooky action of quantum mechanics. That's because they use a non-mainstream method called adiabatic quantum computing.

Unlike classical bits, quantum bits, or qubits, can take the values 0 and 1 at the same time, theoretically offering much faster computing speed. To be truly quantum, the qubits must be linked via the quantum property of entanglement. That's impossible to measure while the device is operating. But in March, two separate tests of the D-Wave device showed indirect evidence for entanglement.

Now Catherine McGeoch of Amherst College, Massachusetts, a consultant to D-Wave, has put their computer through its paces and shown that it can beat regular machines. The D-Wave hardware is designed to solve a particular kind of optimisation problem: minimising the solution of a complicated equation by choosing the values of certain variables. It sounds esoteric, but the problem crops up in many practical applications, such as image recognition and machine learning.

McGeoch and her colleague Cong Wang of Simon Fraser University, in Burnaby, ran the problem on a D-Wave Two computer, which has 439 qubits formed from superconducting niobium loops. They also tried to solve the problem using three leading algorithms running on a high-end desktop computer. The D-Wave machine turned out to be around 3600 times faster than the best conventional algorithm.

Struggling to keep up

McGeoch gave each system roughly half a second to find the best solution to a version of the optimisation problem, and repeated the trial with 100 different versions. She then did the experiment for problems involving even more variables and a more complicated equation.

The D-Wave computer found the best solution every time within half a second. The three regular algorithms struggled to keep up for problems with more than 100 or so variables. The best of the three, CPLEX, had to run for half an hour to match D-Wave's performance on the largest problems.

McGeoch also looked at how the devices tackled two other problems. The D-Wave computer can't solve these directly so employs an extra piece of software to convert them into a form it can handle. D-wave's device had a smaller edge in these cases but still matched or exceeded the capabilities of the algorithms on regular computers. McGeoch will present the results next week at the ACM International Conference on Computing Frontiers in Ischia, Italy.

The number-crunching face-off suggests that the benefits of quantum computers may be harnessed sooner than we imagined. "It's becoming increasingly compelling," says Jeremy O'Brien of the University of Bristol, UK. "If you are a forward-looking business, then you might be that much more compelled to really understand how to use these devices."

It is still possible that D-Wave's device has a strange but highly optimised way of calculating that does not rely on entanglement. Teams will have to repeat the experiments to confirm the quantum effect.

Fairer comparison

The speed tests are also not quite fair, because generic computers will always perform less well than a device dedicated to solving a specific problem, says McGeoch. "A next step would be to build a conventional processor optimised for this task, for a fairer comparison," says O'Brien.

D-Wave's Colin Williams is more certain, pointing out that the company's device finds the best solution in a very different way to regular algorithms. In a classical system, the solutions are poor to begin with but rapidly improve, and then they slowly converge on the best answer. D-Wave's computer reaches the best solution almost instantly. "I've never seen anything like that in a classical algorithm before."

What really matters to the D-Wave team now is convincing customers it has a new kind of device that can help them solve problems more efficiently. "Rather than spending all our time on this academic benchmarking, we would prefer to spend our time on developing real-world applications at this point," says Williams.