D-Wave is also working on a fifth model, which will provide even greater capacity and connectivity and a closer fit to scientists’ needs. It will likely to launch within two years, the machine will again double the number of qubits, to around 4,000. Crucially, it will also provide more-complex connections between qubits, allowing it to tackle more-complicated problems.
“Changing the underlying connectivity is going to be a game-changer,” says Mark Novotny, a physicist at Charles University in Prague, who is exploring a D-Wave machine’s applications to cybersecurity. “I’m basically drooling hoping for it. It’s very exciting.”
D-Wave’s latest 2000 qubit iteration includes an upgrade that Novotny has been clamoring for. The feature gives more control when different groups of qubits go through the annealing process. In at least one case, D-Wave has shown that this can speed up certain calculations 1,000-fold. For Novotny, the feature is crucial because it will allow his team to “sample” qubits during the process, which opens the door to D-Wave exploring a different type of machine-learning algorithm that could learn to recognize much more complex patterns of cyberattacks.
A recent Google study [Phys. Rev. X, 6:031015 (2016)] compared a D-Wave 2X quantum processing unit (QPU) to two classical Monte Carlo algorithms: simulated annealing (SA) and quantum Monte Carlo (QMC). The study showed the D-Wave 2X to be up to 100 million times faster than the classical algorithms. The Google inputs are designed to demonstrate the value of collective multiqubit tunneling, a resource that is available to D-Wave QPUs but not to simulated annealing. But the computational hardness in these inputs is highly localized in gadgets, with only a small amount of complexity coming from global interactions, meaning that the relevance to real-world problems is limited. In this study we provide a new synthetic problem class that addresses the limitations of the Google inputs while retaining their strengths. We use simple clusters instead of more complex gadgets and more emphasis is placed on creating computational hardness through global interactions like those seen in interesting real-world inputs.
We use these inputs to evaluate the new 2000-qubit D-Wave QPU. We include the HFS algorithm—the best performer in a broader analysis of Google inputs—and we include state-of-the-art GPU implementations of SA and QMC. The D-Wave QPU solidly outperforms the software solvers: when we consider pure annealing time (computation time), the D-Wave QPU reaches ground states up to 2600 times faster than the competition. In the task of zero-temperature Boltzmann sampling from challenging multimodal inputs, the D-Wave QPU holds a similar advantage and does not see significant performance degradation due to quantum sampling bias.
But researchers want greater connectivity. Currently, each qubit in the processor can ‘talk’ to only six others, says Scott Pakin, a computer scientist and D-Wave scientific and technical lead at the Los Alamos National Laboratory in New Mexico, which has had a D-Wave computer since August. “The richer the connections, the easier and faster it is to get problems onto the D-Wave. So that’s top of my wish list.”
D-Wave is redesigning its fifth processor to increase connectivity significantly, says Jeremy Hilton, the company’s senior vice-president responsible for technology. And because this upgrade involves a hardware overhaul, it will have an additional benefit: allowing the firm to expand beyond the 10,000-qubit limit imposed by the current processor’s design in future machines, he adds.
D-wave machines are a long way from showing the exponential speed increase over classical computers that their advocates hope to see. But in a paper posted on 17 January and not yet peer-reviewed, a D-Wave team claimed the 2000Q could find solutions up to 2,600 times faster than any known classical algorithm
D-wave’s qubits are much easier to build than the equivalent in more traditional quantum computers, but their quantum states are also more fragile, and their manipulation less precise. So although scientists now agree that D-wave devices do use quantum phenomena in their calculations, some doubt that they can ever be used to solve real-world problems exponentially faster than classical computers — however many qubits are clubbed together, and whatever their configuration. The uncertainty hasn’t stopped the number of users growing: last September, around 100 scientists attended D-Wave’s first users’ conference in Santa Fe, New Mexico.
Existing D-Wave computers are located in the United States, but researchers globally can access them remotely, including through schemes such as the USRA’s. The machines are attracting new kinds of researcher, says Venturelli, who uses one of them to try to find the best way for rovers to autonomously schedule operations and manage time. “Universities with nothing to do with quantum physics are now trying their algorithms,” he says.
D-Wave machines have attracted scepticism as well as excitement since they went on sale six years ago. So far, researchers have proved that, for a problem crafted to suit the machine’s abilities, the quantum computer can offer a huge increase in processing speed over a classical version of an algorithm (V. S. Denchev et al. Phys. Rev. X 6,031015; 2016). But the computers do not beat every classical algorithm, and no one has found a problem for which they outperform all classical rivals.
Quantum annealing (QA) has been proposed as a quantum enhanced optimization heuristic exploiting tunneling. Here, we demonstrate how finite-range tunneling can provide considerable computational advantage. For a crafted problem designed to have tall and narrow energy barriers separating local minima, the D-Wave 2X quantum annealer achieves significant runtime advantages relative to simulated annealing (SA). For instances with 945 variables, this results in a time-to-99%-success-probability that is ∼10^8 times faster than SA running on a single processor core. We also compare physical QA with the quantum Monte Carlo algorithm, an algorithm that emulates quantum tunneling on classical processors. We observe a substantial constant overhead against physical QA: D-Wave 2X again runs up to
∼10^8 times faster than an optimized implementation of the quantum Monte Carlo algorithm on a single core. We note that there exist heuristic classical algorithms that can solve most instances of Chimera structured problems in a time scale comparable to the D-Wave 2X. However, it is well known that such solvers will become ineffective for sufficiently dense connectivity graphs. To investigate whether finite-range tunneling will also confer an advantage for problems of practical interest, we conduct numerical studies on binary optimization problems that cannot yet be represented on quantum hardware. For random instances of the number partitioning problem, we find numerically that algorithms designed to simulate QA scale better than SA. We discuss the implications of these findings for the design of next-generation quantum annealers.