Quantum computer not as fast as maker D-Wave Systems claims, study shows
Lockheed Martin bought the first one in 2011. Two years later, Google, Nasa and the non-profit Universities Space Research Association followed, pitching in together to launch a lab with their new US$10 million-plus toy.
The machines in question are built by D-Wave Systems, the world's only commercial quantum computer company, and are clearly a hot-ticket item.
But the company's bold claim that the computer is thousands of times faster than a conventional PC have ruffled the feathers of academic quantum computing experts, who have been highly sceptical.
Now, new results may give weight to those doubts. An independent research group has found that its second-generation machine, D-Wave Two, shows no evidence of quantum speed-up, a measure of how well a quantum computer is outperforming a conventional PC.
The study was published online on Thursday in the journal Science. It pitted Lockheed Martin's D-Wave system against a conventional PC, posing each a set of problems that gradually increased in difficulty.
"When we just looked at the times - the boring question, basically -for some special problems, D-Wave was 10 times faster," said study author and physicist Professor Matthias Troyer of ETH Zurich. "For other problems, D-Wave was 100 times slower."
But the bigger question of interest is how it will perform in the future, as the machine gets more and more upgrades. To answer this, researchers extrapolated for a greater number of quantum bits, or qubits, and looked for evidence of quantum speed-up.
Basically, this is a measure of how much faster a quantum computer can solve a problem as compared to a conventional one as the problems get harder.
The team saw no traces of quantum speed-up for D-Wave Two. But Troyer said that there could be other types of problems that would show a quantum speed-up, or that perhaps some further fixes on the computer could help.
Jeremy Hilton, D-Wave's vice-president of processor development, said the type of problem that Troyer used for benchmarking was not one for which you would expect an advantage from adding quantumness.
"It's important to consider multiple kinds of benchmarks," Hilton said. "The problem type used [by Troyer] is really not the right problem type."
The new findings are a far cry from previous test results, conducted by a consultant hired by D-Wave, which found that its machine was operating 3,600 times faster than a conventional algorithm running on a PC. The result caused an initial wave of praise by the media, but it was followed by a quick backlash of scepticism by quantum computing experts.
"People started to realise it was just a meaningless comparison," said theoretical computer scientist Professor Scott Aaronson of the Massachusetts Institute of Technology. "When you make the comparison fair, the performance advantage for D-Wave essentially evaporates."