HP's faster-than-flash memristor at least TWO years away
Plus: Storage boffins discuss photonic chip comms
Kavli Foundation Roundtable HP memristor-meister Stan Williams has revealed a product launch delay – saying commercial kit would be available by 2014 at the earliest – and said processor chips would eventually use nanoscale light networking.
He was speaking at a Kavli Foundation Roundtable – the Kavli Foundation being one of those US charitable thingies set up by exceedingly rich business people who have sold their business and need something worthwhile to do that will outlast them. In this case Norwegian-born physicist Fred Kavli sold his Moorpark, California, sensor firm – Kavlico – in 2000 and set up the eponymous foundation, which carries out "an international program of research institutes, professorships, and symposia in the fields of astrophysics, nanoscience, neuroscience, and theoretical physics as well as [awarding] prizes in the fields of astrophysics, nanoscience, and neuroscience."
The foundation bought three boffins together to discuss how atomic scale devices are transforming electronics, and senior HP Labs fellow Stan Williams, popularly regarded as one of the fathers of memristor, was one of them.
Communication is done by photons, computation by electrons and storage by ions
Previously he has said that HP and fab partner Hynix would launch a memristor product in the summer of 2013. At the Kavli do, Williams said: "In terms of commercialisation, we'll have something technologically viable by the end of next year."
But that doesn't mean a commercial product launch, and Hynix's concerns about memristor device effect on flash are relevant: "Our partner, Hynix, is a major producer of flash memory, and memristors will cannibalise its existing business by replacing some flash memory with a different technology. So the way we time the introduction of memristors turns out to be important. There's a lot more money being spent on understanding and modeling the market than on any of the research," said Williams.
We might see a memristor product by summer 2014 but it could be later, as Hynix balances memristor device revenues, starting from zero, cutting into flash revenues in the millions of dollars.
Williams said HP does not want to make memristor chips itself: "Our motivation is not to make memristors, but to ensure there is a supply chain. We're not going to make money off these chips. We are going to make money by building cool systems utilising these chips. Hopefully, we'll build cooler systems than other people because we've been thinking about this technology longer."
He downgraded the importance of Moore's Law: "People talk about reaching the end of Moore's Law, but really, it's irrelevant. Transistors are not a rate-limiting factor in today's computers. We could improve transistors by factor of 1,000 and it would have no impact on the modern computer. The rate-limiting parts are how you store and move information. These are visible targets and we know what we have to do to get there. We can continue to improve data centres and computers at Moore's Law rates – doubling performance every 18 months – for at least another 20 years without getting into something like quantum or neuronal computing."
'We're not going to make money off these chips. We are going to make money by building cool systems utilising these chips'
The Kavli roundtable boffins also discussed light communications in chips. Paul Weiss, Kavli Professor at UCLA and Director of the California NanoSystems Institute, said: "Light is a much faster way to communicate along a device than electrons. But the wavelength of light is much larger than the semiconductor structures we make now. People are used to moving light around with large-scale mirrors or splitters, things we lay out on an optical table. Investigators have come up with very clever ways to do that with nanoscale structures. If we can cut down the size of these structures by one, two, or more orders of magnitude, things start getting exciting about the speed and amount of information we can move on a chip or between chips."
Williams agreed: "At the data centre scale, any time you want to move data more than a few millimetres, you do that as a photon. Our model is that communication is done by photons, computation by electrons, and storage by ions. Each has its limits. Ultimately, they will all be integrated together on a chip." ®