Dwave Systems Tour and interview with CTO Geordie Rose

I, Brian Wang, had a guided tour of Dwave Systems. This was courteously provided by Dwave Systems CTO Geordie Rose.

Highlights

The papers that they publish are about 2 years behind the research and work that they are performing.

They had published a paper back in May, 2011 which provided evidence that the 4 by 4 unit cells leveraged quantum effects. There is no difference between the coupling of the 8 qubit cells with the surrounding cells and recent internal work has been showing that the quantum effects extend to the entire chip.

ie. The full 128 qubits, or 512 qubits (for the new chip) should fully leverage quantum effects.

Dwave has developed adiabatic versions of quantum factoring algorithms. They have an adiabatic alternative to Shor’s algorithm and have factored far larger numbers on their quantum computer system.

The next 18 months will be a critical period for Dwave systems. Raising private money has become far more difficult in the current economic conditions. If Dwave were profitable, then they could IPO. If Dwave were not able to become profitable and IPO and could not raise private capital, then there would be the risk of having to shutdown.

It takes 1 month and dozens of steps (even after automation) to get any of the chips tuned up and ready for use.

One application of the Dwave system is for the optimization problem of creating treatment plans for cancer radiation treatment based on a 3D body scan. This treatment plan takes 1 week using the 128 qubit system but minutes with the 512 qubit system. The optimization algorithm runs 1000 times faster at 512 qubits versus 128 qubits. Cancer radiation treatment plans normally have one person developing the treatment plan and uses a $20 million machine to deliver the precision radiation. The Dwave optimized treatment plan would boost the safety of the treatment by several percent.

The amount of speedup depends upon the different quantum algorithms that are being run.

If the treatment optimization problem were indicative of the speedup of different algorithms, then one might expect (I am extrapolating the 512/128 qubit example)

512 qubits 1000 times faster than 128 qubits
2048 qubits 1000 times faster than 512 qubits

I will have more from the tour and interview in follow up articles.

Cryocooling

Pulse Tube Cryocooler are used, which are described in this 14 page presentation The Pulse tube cryocooler recycles the helium and lowers the operating costs. The cryocooler can achieve tempteratures of 20 millikelvins. The Dwave quantum computer actually seems to work best at 40 millikelvins. A little noise at that temperature actually helps on the hardest problems. On easier problems the temperature does not matter.

Pulse Tube Cryocooler have ~0.5 Watts of cooling at 4K
•Dilution fridge mixing chamber (mK stage) rises ~2 mK/microW
•Pulse Tube fridge requires ~15 kW supply power
–10^10 ratio of fridge supply power to low temp cooling power

If you liked this article, please give it a quick review on ycombinator or StumbleUpon. Thanks