HPC blog I jumped at the chance to interview supercomputing pioneer Bo Ewald and quantum computing whiz kid Murray Thom a few months ago. Although it's been in my “vault of lost content” for a while, the video serves as a good primer for quantum computing and its promise.
It turns out that there are three broad categories of problem where your best bet is a quantum computer. The first is a Monte Carlo type simulation, the second is machine learning, and the third is optimization problems that would drive a regular computer nuts – or, at least, take a long time for it to process.
An example of this type of optimization problem is this: Consider the approximately 2,000 professional hockey players in North America. Your task is to select the very best starting line-up from that roster of guys.
There are a lot of variables to consider. First there's all the individual stats, like how well they score, pass, and defend. But since hockey is a team sport, you also have to consider how well they work when combined with other specific players. When you start adding variables like this, the problem gets exponentially more difficult to solve.
But it's right up the alley of a quantum computer. A D Wave system would consider all of the possible solutions at the same time, then collapse down to the optimal set of player. It's more complicated than I'm making out, of course, but it's a good layman-like example.
So how much faster can quantum computers perform than their digital counterparts? Before purchasing their own D Wave system a few years back, Google put it through its paces and found that when the problem size got to the 500 qubit size range, the D Wave system outperformed its binary cousins by 10,000 times – a solid win in anyone's book.
More recently, Google and NASA found that a D Wave 2 system with 1,097 qubits outperformed existing supercomputers by more than 3,600 times (and personal computers by 100 million x) on an optimization problem, solving it in mere seconds.
I need to point out that we shouldn't get out over our skis when it comes to quantum computing. These performance numbers I'm citing are from corner cases that hit right on the quantum sweet spot. General computing use performance still lags and programming one of these beasts is still more art than science.
Still, I'm intrigued, and want to learn more about quantum computing and the D Wave boxes in particular. Is there RAID on the motherboard? How about PCIe 3? I actually did ask the RAID question in the video – and Bo didn't slap me.