This is the second part of my write-up on my recent visit to D-Wave. The first one can be found here.
The recent shut-down of the US government had wide spread repercussions. One of the side-effects was that NASA had to stop all non-essential activities and this included quantum computing. So the venture which, in cooperation with Google, jointly operates a D-Wave machine was left in limbo for a while. Fortunately, this was short lived enough to hopefully not have any lasting adverse effects. At any rate, maybe it freed up some time to produce a QC mod for Minecraft and the following high level and very artsy Google video that ‘explains’ why they want to do quantum computing in the first place.
If you haven’t been raised on MTV music videos and find rapid succession sub-second cuts migraine inducing (at about the 2:30 mark things settle down a bit), you may want to skip it. So here’s the synopsis (Spoiler alert). The short version of what motivates Google in this endeavor, to paraphrase their own words: We research quantum computing, because we must.
In other news, D-Wave recently transferred its foundry process to a new location, partnering with Cypress Semiconductor Corp, a reminder that D-Wave firmly raised the production of superconducting Niobium circuitry to a new industrial-scale level. Given these new capabilities, it may not be a coincidence that the NSA has recently announced its intention to fund research into super-conducting computing. Depending on how they define “small-scale” the D-Wave machine should already fall into the description of the solicitation bid, which aspires to the following …
“… to demonstrate a small-scale computer based on superconducting logic and cryogenic memory that is energy efficient, scalable, and able to solve interesting problems.”
… although it is fair to assume this program is aimed at classical computing. Prototypes for such chips have been already researched and look rather impressive (direct link to paper). They are using the same chip material and circuitry (Josephson junctions) as D-Wave, so it is not a stretch to consider that industrial scale production of those more conventional chips can immediately benefit from the foundry process know-how that D-Wave has accumulated. It doesn’t seem too much of a stretch to imagine that D-Wave may expand into this market space.
When putting the question to D-Wave’s CTO Geordie Rose, he certainly took some pride in his company’s manufacturing expertise. He stressed that, before D-Wave, nobody was able to scale superconducting VLSI chip production, so this now opens up many additional opportunities. He pointed out that one could, for instance, make an immediate business case for a high through-put router based on this technology, but given the many venues open for growth he stressed the need to chose wisely.
The capacity of the D-Wave fridges are certainly so that they could accommodate more super-conducting hardware. Starting with the Vesuvius chip generation, measurement heat is now generated far away from the chip. Having several in close proximity should therefore not disturb the thermal equilibrium at the core. Geordie considers deploying stacks of quantum chips so that thousands could work in parallel, since they are currently just throwing away a lot of perfectly good chips that come off a wafer. This may eventually necessitate larger cooling units than the current ones that draw 16KW. This approach certainly could make a lot of sense for a hosting model where processing time is rented out to several customers in parallel.
One attractive feature that I pointed out was that if you had classical logic within the box, you’d eliminate a potential bottleneck that could occur if rapid reinitialization and read out of the quantum chip is required, and it would also potentially open the possibility for direct optical interconnects between chips. Geordie seemed to like this idea. One of the challenges to make the current wired design work, was to design high efficiency low pass filters to bring the noise level in these connectors down to an acceptable level. So, in a sense, an optical interconnect could reduce complexity, but clearly would also require some additional research effort to bring down the heat signature of such an optical transmission.
This triggered an interesting, and somewhat ironic, observation on the challenges of managing an extremely creative group of people. Geordie pointed out that he has to think carefully about what to focus his team on, because an attractive side project e.g. ‘adiabatic’ optical interconnects, could prove to be so interesting to many team members that they’d gravitate towards working on this rather than keeping their focus on the other work at hand.
Some other managerial headaches stem from the rapid development cycles. For instance, Geordie would like to develop some training program that will allow a customer’s technical staff to be quickly brought up to speed. But by the time such a program is fully developed, chances are a new chip generation will be ready and necessitate a rewrite of any training material.
Some of D-Wave’s challenges are typical for high tech start ups, others specific to D-Wave. My next, and final, installment will focus on Geordie’s approach to managing these growing pains.