Intel claims optical computing "milestone"
Intel's hybrid silicon laser research has allowed it to transfer 50Gb/s down a single fibre-optic channel.
The milestone - which is part of the company's investigation into the use of optical, rather than electrical, data carriers within computers - is an important one: Intel claims that its latest prototype device is capable of shifting data at a quite incredible 50Gb/s - "the equivalent of an entire HD movie being transmitted every second."
While high-speed optical data transfer isn't new, Intel's work ditches the commonly used 'exotic' materials often used in laser diodes in favour of the cheap silicon it is so used to working with - reducing the environmental impact of the technology while, the company claims, dropping the cost and size low enough for use inside a computer, rather than for external network connections.
Described by Intel as a "concept vehicle" rather than a ready-to-roll product, the system achieves its impressive speeds by combining four silicon-based lasers each carrying a 12.5Gb/s data stream along a single optical fibre, at the end of which the beam is split once again and sent to four photodetectors for decoding for a total of 50Gb/s along a single fibre.
Intel's Justin Rattner, chief technology officer and director of Intel Labs, claims that "this achievement of the world’s first 50Gb/s silicon photonics link with integrated hybrid silicon lasers marks a significant achievement in our long term vision of 'siliconising' photonics and bringing high bandwidth, low cost optical communications in and around future PCs, servers, and consumer devices."
Although clearly part of the same overall strategy as Intel's already-announced Light Peak technology, the company claims that this work is separate - and given that it has already bested Light Peak by a factor of five, the future is looking bright for optical computing.
Are you impressed with Intel's work one this project, or will it take an actual product - available for less than a second mortgage - before you'll start getting excited at the possibilities? Share your thoughts over in the forums.
Previous Article
26 Comments
Discuss in the forums ReplyImagine if a pci-e slot had four of these optical lanes available.. and then imagine a gfx card that would saturate that bus. Like, woah.
That's a good pun, cool tech btw
Then imagine the cost of making a game which would actually stress the said card:)
CGi (film) type graphics/physics... ymmmm! Still, substance over aesthetics - in this context anyway.
Then imagine the cost of buying this Messiah of games and a computer to run it.
And the disappointment when it turns out to have little going for it other than the pretty visuals.
At least Japanese Crysis had the Wakamoto Suit.
It's an efficient miniaturization of optical communication.
All of the number crunching in these is still being done by electrons in silicon. The actual computing is done with electrical transistors. An electronic processor is cruching bits, then this thing is turning those bits into photon pulses, and then the photon pulses are being turned back into electrical bits, where another electronic processor is crunching them.
An "optical computer" is one where the actual processors operate with light instead of electricity. It's a concept that is still far away. The nature of semiconductors allows electricity to control electricity, creating logic gates. As of 2010, we still haven't found a similar device that would allow light to control light. There's a huge amount of research being done in that field.
But it's a misnomer to call this "optical computing". it's a good technology, but it is an optical communication interface, not an optical logic processor.
It is cool to see miniaturization, and I wouldn't mind a link to a technical paper or something. I'm a little confused about the wording of a "silicon". Generally, fiber optics use fused silica. Is this fused silica (which is made from silicon), or some other silicon material? I'm also curious as to what lasers they are using, and what the wavelength is. As far as I knew, the current best technology available is fused silica fibers with semiconducter near-IR lasers at the dispersion minima at 1315nm and 1550nm. if this is a four channel device, it's probably around 1550nm, and is using different bands. One laser might be 1530nm, another 1540nm, another 1550nm, an the last at 1560nm. The other channels are not as efficient as 1550nm, but dispersion doesn't become a limiting factor until these links are several miles long, and we already have big bulky OC fiber for that. This small fiber communications modules would probably be made to make the network infrastructure completely fiber. Right now, we got fiber from the internet cloud to your doorstep, but then most home and businesses have copper from the PCs to the fiber switches in each building.
-They can allways send me their old Xeon's xD
Some nice advancments so far, looking forward to seeing how soon realisticaly this could end up inside all of our home computers ...That bandwidth is huge!!! :)
you can make the boards lighter and more energy efficient, batteries will last longer, as you won't be losing power along the tracks of the motherboard. essentially, you would have a power bus, maybe just +5v to provide power only to the modules, all communication, between modules would be optical, not electrical.
you could in theory get rid of the pci/pcie bus, and have a 2 wire connector to provide power only, with an optical connector for the data communication. i think that is the kind of goal they are looking at.
Is it a silicon and gallium arsenide hybrid? Or all silicon. I didn't think an all silicon laser would work, unless it is something really nifty like a quantum cascade, but those are too far in the IR to be good over fiber.
yeah but how does it work? Silicon is an indirect bandgap, so you can't make a semiconductor laser out of it like you can from gallium arsenide. I can double check my good friends Saleh and Teich when I get home, but I don't recall it being possible.
Theoretically, they could make a quantum cascade laser out of silicon, but I have never heard of that being done, and typical QCLs are way outside the communications band.
I think they are wording it wrong. It's gotta be gallium arsenide fabricated inside a silicon wavegide, where the actual laser medium is still gallium arsenide. I know Intel has been doing research in that. But all silicon? I have my doubts, but I'm very curious now for further info.
as for optical links on the motherboards itself Intel would have to develop some way to print optical fibers onto the PCB or gluing on fiber to the pcb somehow. Guess robots can handle everything these days.