Slashdot Mirror
New Material Sets Stage For All-Optical Computing
An anonymous reader writes with this excerpt from the International Business Times: "Researchers have made a new material that can be used to guide waves of light, a breakthrough that could lead to ultra-fast computing. Georgia Tech scientists are using specially designed organic dyes that can process and redirect light without the need to be converted to electricity first. ... 'For this class of molecules, we can with a high degree of reliability predict where the molecules will have both large optical nonlinearities and low two-photon absorption,' said [Georgia Tech School of Chemistry professor Seth] Marder."
According to the article, using an optical router could lead to transmission speeds as high as 2,000 gigabits per second, five times faster than current technology.
"2000 gigabits per second"
GigaBITs? Wow!
"five times faster than current technology." Reminds me of being a teenager and discovering lotion...
I could not get my eyes from that advertisy pic.
Must...read...article.
EETimes has "IBM Research claimed a keystone achievement in on-chip optical communications Wednesday (March 3), saying its 40-gigabit-per-second (Gbps) germanium avalanche photodetector completes what it calls the nanophotonic toolkit." (link) (A few days before announcing 2,500 layoffs, hmmm...)
...And the same news from Semiconductor Intl.
I'd like to benchmark this against graphene. Since optical signals don't have to be converted to electrical first, then (I think) the bottleneck would be the optoelectronics.
Entomologically speaking, the spider is not a bug, it's a feature.
Yes. This keeps the data in "optical form", which has additional ramifications. Once (if?) this becomes practical, additional uses will crop up.
Entomologically speaking, the spider is not a bug, it's a feature.
Sure -- its for routers, a basic component of data transmission, and a bit is the most fundamental piece of data. We typically think in "BYTES", but that's really an OS abstraction right? At the lowest level of the network stack it makes sense to talk in bits -- anything above that can interpret it as it will (like a Byte, Word, Long.. etc)
Exactly, like lower latency. The conversion into an electrical signal, and then back to optical probably adds a bit of latency. I'm no expert, but I'd imagine that data going to and from a typical destination on the internet goes through several of these conversions adding (in most cases negligible) latency. If most of the routers on the net were all optical, I'd imagine we'd have an internet with imperceptible latency most of the time. That could lead to things as simple as lag-free gaming, real-time video conferencing, and maybe in the future a very (sur)real shared virtual reality, all done across large physical distances.
Of course. Because the new technology also is getting better. And usually at a much quicker rate than the existing one, because that one is already at the end of its limits.
There often even is new technology that is still worse than the old one, because of its experimental state. But worth pursuing anyway, because of the huge potential.
The same is true for optical circuits.
Any sufficiently advanced intelligence is indistinguishable from stupidity.
If the best expected performance of the new technology is just 5 times better than current technology, is it really worth pursuing it?
Moore's law says 2× in two years, but some people think we're running up against the limits of Silicon.
This gives us 5× in one shot.
You ask is it worth it? Are you suggesting that we not engage in pure science any more because it might not pan out in the long run? Only time will tell if it was worth it.
I would consider 3 factors: "need", "current capacity", and "future capacity". "Need" is always increasing, at a rate to overtake "current capacity," therefore "future capacity" needs to always be developed. If they are predicting 5x faster performance at this early stage, it surely holds more. When BluRay was in it's infancy many people said 'why would i ever need 50GB, thats bigger than my HD'--- but low and behold, now 50gb doesn't seem like enough
+1 for Georgia Tech, twice in one week (Spanish botnet taken down)
Where did you learn it was an OS abstraction? That's just ... sigh.
Bytes are the smallest addressable unit of memory the CPU can handle. It doesn't matter if the memory controller only does cache line fills or whatever, memory addresses are in units of bytes.
The first automobiles could easily be outrun by a horse. I guess we're fortunate that no one noticed that or else they would've all agreed that automobile technology was a waste of time and should be abandoned.
One time I threw a brick at a duck.
I've been reading headlines for the past 20 years that claim "breakthroughs" in all-photonic computing. Where are the all-photonic routers?
This is a result of the highly-clustered, highly-mobile computing age we live in today. A single fast chip isn't as applicable any more. Give us tiny and low-power.
The big issues in designing optical switches is their switching time and minimum switch pulse width. I and my group built what is probably the first all-optical computer in the early '90s. We used Lithium Niobate switches, which limited the machine's clock frequency of 100 MHz. It's hard to find the original article, which is in the Feb. 18 issue of Science Express. Subscription required, unfortunately. In that article the authors say nothing about switching time, or minimum switch pulse. It looks like a good piece of research, but eons away from anything practical.
It takes light 64ms to go from NYC to Beijing, as the crow flies. So roundtrip of 128 ms is borderline unplayable in first person shooters
Obviously the ramifications as far as emissions security (TEMPEST though that's a simplification of TEMPEST) are huge, but what is this likely to do for heating and component size. I can see this being a great opportunity for a lot of military applications even if the speed limit is only a few times better than what we have now.
"Growing old is inevitable; growing up is optional."
It's my understanding that fiber optic cable has speeds that are limited by the electronic conversions on either side. Is that what is the issue here as well? How well could this (for lack of a better term) internal light mesh with an external (fiber optic) light?
"A government is a body of people usually -- notably -- ungoverned." -Shepherd Book
Unfortunately, the article didn't hint at this possibility at all. However, I did pick this up:
The research was funded by the National Science Foundation (NSF), the Defense Advanced Research Projects Agency (DARPA) and the Office of Naval Research (ONR).
So DARPA's helping fund it eh? In answer to my own question then, "Yes!"
Leave it to DARPA to fund the development of a cloaking device and play it off as a computer breakthrough. I, for one, am stoked.
Motorcycles, Robots, Space Gossip and More!
Just a guess, but i wouldn't imagine packets are measure in bits but in bytes. That's why hex is often used, right? Unless the router is using an exotic protocol, everything will be measured in bytes? Data structures are almost all in multiples of bytes.. size_of(bool) returns 1. malloc works with bytes, not bits. I'm pretty sure you will have a hard time finding anything that works directly in bits and not bytes, to include routers.
http://soylentnews.org/~tibman
I'm a gamer you insensitive clod!
Nice car analogy!
Smoke and mirrors: A new material that can be used to guide waves of light
Keep Doing Good.