D-Wave's announcement of the first “commercially viable” quantum computer has been met with a polarized mix of excitement, skepticism and head-scratching confusion. Friendly voices express excited optimism, but tongues are also clicking in the heads of those who say they recognize hyper-hyperbole when they hear it. So, which is it? Have we just heard Krakatoa erupt, or the pinging of a BB that missed its target? And what practical effect does this have on HPC and the broader computer market?
It depends on whom you talk to. A recent survey of over 700 IEEE Fellows showed that seventy-eight percent of respondents had doubts that a useful quantum computer will reach the market in the next 50 years. Will this announcement change that sentiment? Stay tuned.
The quantum computer's main promise is to solve NP-complete problems in polynomial time. These problems cannot be obtained or would take too long on a silicon supercomputer, no matter how large. Such problems are addressed today through techniques such as heuristics, often at unsatisfying, small scales. When a rough guess is the best available, a natural market opportunity emerges for QCs. While there is plenty of work to go around for such problems, QC seems destined to remain exotic and rare for years to come.
To get a sharper view of the potential impact of quantum computers, avoid being distracted by the physics that is at once the glory and the source of confusion about this technology. Focus on the key question; what does commercially viable mean? If it requires that quantum computers made from superconducting niobium circuits, chilled to something like absolute zero take a principal place in data centers and solve our toughest problems, then it might indeed take 50 years. In fact, there are so many variables inherent in that question that we may need a working quantum computer to answer it with certainty. However, if it means low quantities will be sold at high prices to solve specific problems for those few with nearly-unlimited funds, then today's demonstration of the D-Wave system may indeed be significant to the wizards behind the curtain of high performance computing.
To understand QC, major barriers have to be overcome by the press and computer scientists alike. Why? Because QC exists in the bizarre world that is the nexus of quantum physics, mathematics, computer science and science fiction. Few accessible sources exist to explain this phenomenon to the likes of this reporter whose 27 years in the computer business did not (yet) include enough quantum physics. We need more than MIT papers and the few blogs shared by the tiny crowd of QC cognoscenti — a very limited list of names, institutions and opinions. Mind you, niobium, Cooper pairs, absolute zero and the memoirs of Lord Kelvin make a great hobby, but there's only so much time on weekends.
At what point, if ever, will quantum computing be competitive with the silicon/hafnium and the “teraflops-on-a-chip” architectures we have learned about over the last few weeks? Most sources seem to say “never,” typifying QC as a sort of specialized computer, comparable to the role of the original vector supercomputers. When will development tools and languages become available to allow scientists to state NP-complete problems? How will the system scale? Ready answers are not available yet, but D-Wave has chosen to position itself as the leader, and with that position comes an excellent opportunity, perhaps a responsibility, to educate and clear the way to acceptance.
What might the scale of the market be? This sort of long-term specialization is remindful of two current technologies that occupy small but secure market niches — vector processors and FPGAs. Cray, too set about solving simulation problems with exotic technology once, cooled by Freon. FPGAs are growing in response to specialized problems with a high-stakes need for speed, and their market potential may be considerable. But neither of these technologies can be described as potentially disruptive to mainstream computing, nor even mainstream technical computing.
This suggests a substantial period where QCs should be thought of as specialized computers with important but limited application. In the absence of data and research, perhaps the history of vector processors gives us a hint to market potential for the near term, given reliable hardware and tools that scale to solve pragmatic problems. NEC hoped to sell 700 SX-8 units over a period roughly from 2004 to 2008. Cray is seeking to revitalize their vector business with adaptive combinations of compatible vector and scalar systems, but by design unit sales are likely to be in modest numbers.
Will systems like the D-Wave adiabatic quantum computer shape architectural decisions at Intel, IBM, Sun and AMD? D-Wave makes no such claim but hints at a go-to-market strategy that would include larger players. Kudos to D-Wave for using the power of PR to put the topic on the radar, for gaining mindshare at some risk, and for taking a bold step to lead the market and bear the burden of proof. It may not be valuable to ask how long it will be before current parallel architecture moves aside and the deep cooling of superconducting Josephson-style circuits replaces the heat of silicon systems as a problem in large data centers. Maybe 50 years is not so far fetched. On the other hand, we may all be in for one heck of a physics lesson.
Bob Feldman is president of HPC Marketing (www.hpcmarketing.com), a Sales and Marketing Consultancy for Advanced Technology Companies, based in Harvard Massachusetts, and a contributor to HPCwire.