Bleeding Edge: Scalable Optical Processing
August 18, 2009
Filed Under: in Analysis, Bleeding Edge, Infrastructure 2.0, News
Author: Mark 'Rizzn' Hopkins
Welcome, new visitor. If you haven't already, you should create an account, subscribe to our RSS feed or perhaps follow us on Twitter. Thanks for visiting!
(creating an account will prevent this message from showing up again)
If you're a tech blogger, create an account and ask an Admin about our free cloud blog hosting, powered by Rackspace.
More from the “Moore’s Law Can’t Die” department today from Ars Technica’s Chris Lee:
One of the main barriers to reducing the size of optical components is the wavelength of light. Visible light has a wavelength of around
500nm, so devices that manipulate light, like lenses and waveguides, must have comparable sizes. At least up until now—a long-awaited development has now provided a proof of principle, demonstrating that lasers can be made as small as 50nm, sizes that are comparable to current electronic features.
So, we now have 50nm laser hardware, which could conceivably be combined with nanowires to start developing optical circuits that really do look like electronic circuits (e.g., small and cheap). This laser was powered by a very powerful pump laser, meaning that it's only small if you ignore the enormous power supply. But that was a side effect of how the experiment was put together. A single spaser used about 20 microwatts of power, so much smaller pump sources are feasible. If they can achieve continuous wave operation, the researchers are on to a winner.
Yesterday we talked about IBM’s ten year plan for DNA computers – this experiment comes from the science journal Nature, so it’s even further afield than IBM’s DNA computing technology, and is purely theoretical at this point.
It is pretty well thought out and interesting, and represents a breakthrough in optical technology. The bottleneck seems to be power requirements, which until breakthroughs occur within electricity conversion tech, could limit the applications to supercomputer and cloud applications.
It’s something to keep on your radar, and could represent how traditional microchip methodologies could keep pace with the more esoteric emergent computing technology.
I am often confused by what Intel is doing with respect to the future of Moore's Law? If the world turn in this direction this means that Intel will be at risk.
What about other processor companies?
Moore's Law is something of a constant driven by a combination of psychology and market dynamics. There might be a danger that Moore's Law doesn't continue of Intel finally succeeds in thoroughly trouncing AMD, and competition plays a diminished factor in driving their R&D cycle.
Even before Moore's Law and the Law of Accelerating Returns was codified, it has been statistically true, so if Intel fails to perform by that standard, someone else will.