What is Digg?Sponsored by Dragon Age: Origins
Follow the Dragon Age: Origins development team on Twitter view!
twitter.com/DragonAge - EA presents BioWare's new dark fantasy epic Dragon Age: Origins. '9/10' from Game Informer.
25 Comments
- elnerdo, on 10/12/2007, -0/+10Because you didn't use the word 'petaflop'
- inactive, on 10/12/2007, -0/+3Isn't the emmission of the bursts of light still controlled by traditional electronic type stuff?
A digg story from a few days ago about a 1Terrabyte solid state storage system (http://digg.com/hardware/1_Terabyte_of_Solid_State_Hard_Drive_Space_) claimed 24Gb/sec speeds, so what is the point of this laser thingy if we can allready get similar speeds out of existing tech?
Sure, the current chip-to-chip speeds isn't great, and I guess this is where the laser thing comes into it, but to suggest that you can't get anywhere near 25Gb/sec transfers without lasers is just baloney! - Lacero, on 10/12/2007, -0/+3No heat too... just photons zipping around.
- inactive, on 10/12/2007, -1/+3frickin Laser Beams!
- yensed, on 10/12/2007, -0/+2The Future Is Here! Or, Well. Will be in the Future. Which will make it the Present. But not the Now present, but the present of the future... Then every phone call, email, Digg comment and IM's from every state can be viewed simultaneously from 1 super computer.
- locutis, on 10/12/2007, -0/+1Actually when I first read about this technology a few years back the chips were submerged in liquid nitrogen. Maybe they have sorted this out. For synchronous operation size is also an issue, since light takes a leisurely 3 picoseconds to travel a mm.
- everfalling, on 10/12/2007, -0/+1i'm wondering how th light is generated and detected. at least with metal and electrocity there are ways of storing electrocity in chips and the like, all from a single source of power. with optical circutry, i think of things communicating like strobing a flashlight in front of a persons face. one, you need the light generator, then you need the detector, then the path between them. i can see you making optical fiber small enough, but how do you generate the light in such a small space? and wouldn't the detectors of the light still work offa electricity to interperate the strobing? and how do you store the light information? it'd be nice if there was a link to some kinda illustrated or animated guide on how this technology works. i think i'm so used to understanding electricity that it's hard for me to see light being used in such a compact space.
edit: also, i was thinking. would this optical circutry just be in the CPU, or in other things as well? i see hard times ahead for computer modders who can't simply solder wires onto chips anymore. - lookoutforchris, on 10/12/2007, -0/+125Gb/sec = 3.125GB/s. Which is not very fast when compared to a high speed bus in say a graphics card. It's pretty fast for a CPU-Memory interconnect in a large NUMA system. Though Silicon Graphics' interconnect (NUMALink) is good for 3.2GB/s one way (6.4GB/s bidirectional). I think quad rate 12x Infiniband is good for 12GB/s, though I don't know of any system using it (Cray maybe?). Maybe this interconnect will be multi-plexed or aggregated some way? The article doesn't say. Maybe latency of these NEC interconnects is much lower because of the optical fiber? The article doesn't mention.
- fletchowns, on 10/12/2007, -0/+1Great news for fluid dynamics!
- inactive, on 10/12/2007, -0/+1That's funny, in Australia we call them 'laser fries'.
- JQP123, on 10/12/2007, -0/+1Disclaimer: I'm not a physicist either so this is quasi-speculation on my part. High frequency electrical signals are susceptible to electromagnetic interference. To help detect and differentitate between a real signal and "noise", an electrical signal is typically maintained and sampled multiple times over a small but finite time interval. The required duration effectively limits the practical communication speed.
Light is immune to electromagnetic effects and thus the required signal duration can be made shorter. - mesoed, on 10/12/2007, -0/+1It's about time there was experimenting done with optics in computer hardware. While a completely different beast, the telcos figured out optics were better long ago. I guess it was only a matter of time before this was applied at the nanotech level.
- Murdats, on 10/12/2007, -0/+1I have only have basic education in physics but doesnt light travel at the same speed as electricity
the only way i see lasers being able to transmit data faster is by fitting more 'bands' in a smaller space if this is possible. could someone please explain how the lasers make it faster? - sonofagunn, on 10/12/2007, -0/+0@NoMoreNicksLeft
Luxtera has built and demonstrated solid-state lasers in silicon and optical interconnects running at 40 Gbit/s. The cool thing about their's is that it's built entirely on silicon chips using existing manufacturing techniques. - JQP123, on 10/12/2007, -0/+0"Isn't the emmission of the bursts of light still controlled by traditional electronic type stuff?"
Yes, electrical signals are being converted to light and back again. Such conversions typically involve some loss of energy so I wouldn't expect this to be a power saving feature.
".. 1Terrabyte solid state storage system ...claimed 24Gb/sec speeds, so what is the point of this laser thingy if we can allready get similar speeds out of existing tech?"
I'm assuming this article addresses a single serial communication line. You can't really compare this to something like a storage system which most likely uses multiple lines operating in parallel to achieve greater speeds. - mincua, on 10/12/2007, -0/+0that were lots of ddram chips running over a fiber channel you cand not get 24gbps for ome single ddr chip
- NoMoreNicksLeft, on 10/12/2007, -0/+0Though silicon itself isn't very good for it, other semiconductors can actually be made into solid state lasers. They can even "etch" the laser into the substrate itself. The only reason we haven't done it earlier, is that no one wants to move over to indium or gallium... so much is invested in silicon fabs.
- lmlloyd, on 10/12/2007, -0/+0I don't understand why this is really news. AT&T developed optical processors quite a long time ago (I'm talking late '80s, early '90s, back when they still had their Graphics Software Lab). The reason it never went anywhere back then was that they had the processor, but no optical memory. MCC then developed an optical processor and holographic memory in the mid '90s, and it once again never went anywhere, and just stayed in the lab.
Both of these systems had huge potential, but never went anywhere. Why should we think this newest discovery of the same technology is going to be any different? - lmlloyd, on 10/12/2007, -0/+0The thing is that electricity in unidirectional. At any given time on any given path, the current will only run one way. This mean that if you want to send signals both ways, you either have to have two paths, or have to send the signal one direction, then reverse the polarity, and send it the other direction.
Light doesn't work that way. It can pass through itself, and travel both directions on the same path. That means you can simultaneously send signals in both directions on the same path. Further more, you can cram even more data through the same pipe by varying the wavelength. - JQP123, on 10/12/2007, -0/+0"Why should we think this newest discovery of the same technology is going to be any different?"
Goal here is much more modest than a full fledged optical computer. They're simply building an optical bus as a fast way to transfer data between processors and other components within a supercomputer. - RobGamble, on 10/12/2007, -0/+0That's huge, because the average chip fabrication plant costs in the billions, so Motorola, IBM, Intel, etc. cannot simply "retool" a fab every time some new technology comes out.
- JQP123, on 10/12/2007, -0/+0Communication "distance" is a significant factor here. The longer the distance, the more beneficial optical communication techniques become. For the telcos sending signals over long distances, it's a no-brainer but once you get down to the nanotech and nanometer level, I'm not convinced that any advantage remains.
- ElevenisEven, on 10/12/2007, -1/+0petaflop...if you only knew...lol
- VanceMc, on 10/12/2007, -3/+0Duplicate Story: First Posted Here:
http://www.digg.com/technology/Laser_chips_could_power_petaflop_computers - guydoyen, on 05/02/2008, -10/+4I've submitted this story 16 hours ago : http://digg.com/technology/NEC_Laser_communications_chips_can_transfer_information_at_25_Gbps
I don't understand why almost nobody has digg it... Can somebody tell me why ?
© Digg Inc. 2009
— Content created and posted by Digg users is