Intel creates silicon photonic cable capable of 50Gbit/s

[sdsdv10]

Intel has just unveiled a new high-speed fiber optics cable dubbed the Intel 50G Photonics Link, which uses silicon photonics to reach speeds of 50Gbit/s. The new cable uses a hybrid silicon laser and a transmitter chip to split the signal into four 12.5Gb/sec channels, which are combined to create a fast 50Gbit/s connection on a fibre cable as thin as a piece of human hair. Intel has said this new technology could lead to low-cost optical links in the future and even be capable of reaching 1Tb/sec.

http://www.neoseeker.com/news/14453-...le-of-50gbits/

Cool videos here;
http://www.youtube.com/watch?v=0U4Af...eature=channel

http://www.youtube.com/watch?v=U5HhR...eature=channel

EDIT: additional news links

http://www.theregister.co.uk/2010/07...con_photonics/
http://www.wired.com/gadgetlab/2010/...onics-50-gbps/

[hecktic]

awsome when will this be mainstream

[shabby]

awsome when will this be mainstream

In a few light years...

[Darakian]

In a few light years...

You'll get it when you're ready to fight brock! :p

[unrlmth]

In a few light years...

When we start measuring time in light years.

[Serra]

At first I was all but then I was like . Then I read a bit and I was kind of happy, but still not terribly so.

What Intel is really advertising here is 4x 12.5Gb/s links, but we can already do that and more. I could go buy a 100Gb/s link right now if I were inclined and had the cash.... they use 10x 10Gb/s links. The advertisement here is really just that they can hit 50Gb/s in what looks like a smaller footprint.. though truth be told it might just be that there is less sheathing (current fiber optics can have tiny, tiny optical wires... most of the size is in the jacket).

Also, 12.5Gb/s will not be the next speed spec, 40Gb/s is humming along quite smoothly, as is a 100Gb/s spec. So in that way I'm pretty sure they are really only focusing on the hardware here, as creating a 12.5Gb/s spec would be an odd move right now to say the least (ratification time takes forever).


I'm sure there were some lessons learned here that will help shrink processes in the future and drop prices some, but this looks like 95% marketing to me.

[Johnny87au]

awesome mate!

[terrace215]

but this looks like 95% marketing to me.

The breakthrough is the *integrated* *silicon* laser part.

mass-produced silicon "Lasers on die" = cheap lasers.


February 2004: world’s first 1 GHz silicon
modulator
· February 2005: world’s first continuous
wave silicon Raman laser
· April 2005: world’s first 10 Gbps silicon
modulator
· September 2006: world’s first hybrid
silicon laser
· July 2007: world’s first 40 Gbps silicon
modulator
· August 2007: world’s first 40 Gbps PIN
photodetector
· December 2008: world’s first 340 GHz
Gain*BW avalanche photodetector (APD)


etc

http://www.intel.com/pressroom/archi...727comp_sm.htm

[EvilOne]

Miniture lasers

Where's the oversized moon killer laser they should have been working on.

[saaya]

serra, yeah when cables are this thin, you really wonder what the big deal is over having one or 4 of them to get the same speed, right?
i mean yeah, thats cool... and an improvement... but nothing really game changing... i think the real break through is the transciever i guess...
and thats nice... would be a nice way to upgrade, you basically double or quadruple the bw per cable you already have installed...
thats what i think this is about, not really about the bandwidth they achieved...

oh and fyi guys, lightyears is a measurement unit for DISTANCE, not time

[The_Beast]

In a few light years...

When we start measuring time in light years.

unrlmth is right, a light year is a unit of distance not time

[Aberration]

serra, yeah when cables are this thin, you really wonder what the big deal is over having one or 4 of them to get the same speed, right?
i mean yeah, thats cool... and an improvement... but nothing really game changing... i think the real break through is the transciever i guess...
and thats nice... would be a nice way to upgrade, you basically double or quadruple the bw per cable you already have installed...
thats what i think this is about, not really about the bandwidth they achieved...

oh and fyi guys, lightyears is a measurement unit for DISTANCE, not time

It is not so much to do with the size of the fiber but everything else that goes with it. If you use 10 fibers then you need more hardware than for 4.

So in that essence it is about the bandwidth per fiber, because that drastically reduces everything else needed.

[Hornet331]

At first I was all but then I was like . Then I read a bit and I was kind of happy, but still not terribly so.

What Intel is really advertising here is 4x 12.5Gb/s links, but we can already do that and more. I could go buy a 100Gb/s link right now if I were inclined and had the cash....

And thats the important part, current products costs a couble thoused dollar, with that technologie the price can drop to a few dollers, cause its all made on silicon. Imagin routers/switches at consumer prices that deliver that speed.

[Aberration]

And just imagine what this means to the ISPs who have enormous networks to upgrade.

Stuff like this can only mean great speed for the end user at reduced costs, eventually of course.

[sdsdv10]

It is not so much to do with the size of the fiber but everything else that goes with it. If you use 10 fibers then you need more hardware than for 4.

So in that essence it is about the bandwidth per fiber, because that drastically reduces everything else needed.

Just to clarify here, this is 4 wavelengths of light down just one (1) fiber via multiplexing. One of benefits of data transmission via light waves. All you need is additional wavelenghts of light, not additional fibers, to get more bandwidth. As Intel continues to shrink the process, I'm sure they will both increase the data capacity per wavelenght of light and increase the number or channels (or wavelengths) of light per integrated transceiver.

Note, one of the other big benefits here over copper is cable length. One of the videos states they can get the 50Gb/sec on a cable over 100 meters in length without any signal boosting/enhancing required. Since this technology is still very young, I'm sure that will grow as will the data rate.

[saaya]

It is not so much to do with the size of the fiber but everything else that goes with it. If you use 10 fibers then you need more hardware than for 4.

So in that essence it is about the bandwidth per fiber, because that drastically reduces everything else needed.

good point... and while the fibre cable is tiny the protection wrapped around it makes it a lot thicker.. still very thin but i guess for isps that does make a difference if they are trying to cram as many of these through some old sewer pipes

And just imagine what this means to the ISPs who have enormous networks to upgrade.

Stuff like this can only mean great speed for the end user at reduced costs, eventually of course.

whats that? youtube HD videos actually streaming at the bitrate they play at? HALLELUJAH!

[Sn0wm@n]

At first I was all but then I was like . Then I read a bit and I was kind of happy, but still not terribly so.

What Intel is really advertising here is 4x 12.5Gb/s links, but we can already do that and more. I could go buy a 100Gb/s link right now if I were inclined and had the cash.... they use 10x 10Gb/s links. The advertisement here is really just that they can hit 50Gb/s in what looks like a smaller footprint.. though truth be told it might just be that there is less sheathing (current fiber optics can have tiny, tiny optical wires... most of the size is in the jacket).

Also, 12.5Gb/s will not be the next speed spec, 40Gb/s is humming along quite smoothly, as is a 100Gb/s spec. So in that way I'm pretty sure they are really only focusing on the hardware here, as creating a 12.5Gb/s spec would be an odd move right now to say the least (ratification time takes forever).


I'm sure there were some lessons learned here that will help shrink processes in the future and drop prices some, but this looks like 95% marketing to me.

isn't the max 120gb/s for infiniband right now only possible by aggregating a couple of cards together ????

[Serra]

The breakthrough is the *integrated* *silicon* laser part.

mass-produced silicon "Lasers on die" = cheap lasers.

<snip>

If that's the breakthrough I'm surprised their announcement was for the cable:

"the Intel 50G Photonics Link, which uses silicon photonics to reach speeds of 50Gbit/s. The new cable uses a hybrid silicon laser and a transmitter chip to split the signal into four 12.5Gb/sec channels"

With that said, after the bolded part they do point out it uses a hybrid silicon laser, so I guess they could be confusing terminology and do mean to point out the transmitter/receiver rather than the cable... I would accept that argument.

Also, I seem to have misread this last night and it is a 4-color transceiver rather than 4 wires, and that's a notably positive difference. Still, I'm not really excited by it. I'm not saying it's not an advancement and that it doesn't please me, it is and it does. But it's not something I'm excited about, and here's why:

1. In the best-case scenario it's a 5-year technology that will enable us to do what we have done for the last 2 years at a lower price (compared to todays prices)

2. I don't know the exact specs on this, but we do currently mass-produce 2-color 10Gig links (20Gb/s throughput), and should be doubling that soon. Research is currently ongoing for 8-color transceivers elsewhere as well. Definitely not with the same processes, but a 4-color transceiver in a lab isn't exciting for me.

3. The best use for this technology (beyond possibly reducing prices in networking) would be inside the PC, replacing various buses - that idea excites me. But they are still very, very far from that with this prototype announcement. I talk about using this in inter-computer networking as an example because the parts can be bigger and slower in that arena, but inside the PC this is just touching the surface of what needs to be done.

4. Keeping in mind the above there is another challenge. Lasers like this are great for inside a PC - they're small and fast (in this case probably not small or fast enough, but it'll get there eventually)... but there is one hurdle they did not address in their most likely debut area if this is going to be used in the next 5 years - distance. I can transmit 10 Gigs a huge distance on existing equipment, and they demonstrated it on what looks like a 2-3 foot cable. I could barely connect two switches in a rack with a cable that small. If they had pointed out that it could be used over a distance of, say, even 100m I would be considerably more impressed because that would be considerably more useful. I would be dollars to donuts however that it is not a distance technology (yet), and if not that's another X years of research that need to be added on to the clock before a link of even this speed comes out.


That's why I'm not excited. I'm happy (don't get me wrong about that), but not excited. I'll save my excitement for when they prove it can be mass produced or provide some deeper specs that prove it's even useful in its current implementation (see point 4) and do more than butter me with marketing.

isn't the max 120gb/s for infiniband right now only possible by aggregating a couple of cards together ????

Yes, and that's one of the things that makes me happy to see this. But, in fairness, this system would also need at least 3 cards to get to 120Gb/s (I'm guessing 3 would just hit the requirement, because I'm guessing that for marketing purposes they gave the transmit rate as the throughput rate, rather than taking encoding into account... which would mean it's probably a 40Gb/s usable technology). We could do it in about 6 now, but should be able to do it in 3 before too long (at the highest end of the price scale though, likely much more expensive than this system would be).

[Aberration]

As to #4, its fiber optics. I doubt the length they demo'd it on was of any concern, and why they probably did not mention or focus on the distance able to transmit.

As to what they are debuting, I think it is a misprint calling it just the cable. I think it comes from it being called the Photonic Link. They probably saw link and just drew the assumption to the cable.

"This achievement of the world's first 50Gbps silicon photonics link with integrated hybrid silicon lasers marks a significant achievement in our long term vision of ‘siliconizing' photonics and bringing high bandwidth, low cost optical communications in and around future PCs, servers, and consumer devices," said Justin Rattner, Intel chief technology officer and director of Intel Labs.

Its the whole package, not just the cable. From what I can tell there is nothing new in the actual fiber.

Also from the article quotes, they intend this for all types of communications, internal and external and over great distances.

Dont you guys remember the Photonics articles floating around about Intel on XS? This is the culmination of that.

[jam2k]

can I has ftp ?

[Exa.bit]

This reminds me of the light peak technology Intel came up with awhile back.
http://techresearch.intel.com/articles/None/1813.htm

Light Peak is the code-name for a new high-speed optical cable technology designed to connect your electronic devices to each other. Light Peak delivers high bandwidth starting at 10Gb/s with the potential ability to scale to 100Gb/s over the next decade. At 10Gb/s, you could transfer a full-length Blu-Ray movie in less than 30 seconds. Optical technology also allows for smaller connectors and longer, thinner, and more flexible cables than currently possible. Light Peak also has the ability to run multiple protocols simultaneously over a single cable, enabling the technology to connect devices such as peripherals, displays, disk drives, docking stations, and more

Seems Intel is putting a lot of R&D into fiber lately. It sure would be great if some of this tech. goes mainstream soon. Earth is bound to run out of copper eventually, light is a great replacement.

[haylui]

this happen only in the lab
and btw, what kind of storage device that gives you 50Gbit/s output?

[Serra]

As to #4, its fiber optics. I doubt the length they demo'd it on was of any concern, and why they probably did not mention or focus on the distance able to transmit.

Fiber optics don't "just work", there is a lot more to it than that. If that were the case, I could use one laser type to simply run my 10km links or my 10m links... it's just not. To push a signal further the power requirements are higher, and so too are the cooling requirements. Saying it's not an issue to push a signal further is like saying it's trivial to just push your processor to 5x its rated frequency... it may seem that way on the top, but it just isn't the case. One thing to keep in mind is that it's easier to push information on a high frequency wave, but it requires more power. Depending on the wavelengths used, that may (or may not be) a CONSIDERABLE amount more power for longer distances at this same bit rate.

I would hope that it can push distances, but when you're announcing lab achievements that is not always the case and I will wait until I can see a confirmation from Intel before I expect that they can perform at distance. I think that's fair skepticism.

[Sn0wm@n]

this happen only in the lab
and btw, what kind of storage device that gives you 50Gbit/s output?

10Gb ethernet and the later 40Gb ethernet that is comming really soon ,infiniband,sas etc... take your pick ... server oriented storage needs can give you that much bandwith today and more ....

[sdsdv10]

Anand has a great article on this with a lot more pictures and descriptions of the technology involved.
Check it out here: http://www.anandtech.com/show/3834/i...photonics-link