Beta
×

Welcome to the Slashdot Beta site -- learn more here. Use the link in the footer or click here to return to the Classic version of Slashdot.

Thank you!

Before you choose to head back to the Classic look of the site, we'd appreciate it if you share your thoughts on the Beta; your feedback is what drives our ongoing development.

Beta is different and we value you taking the time to try it out. Please take a look at the changes we've made in Beta and  learn more about it. Thanks for reading, and for making the site better!

Intel Demos Optical Data Transfer For Servers

samzenpus posted about a year and a half ago | from the beam-of-light dept.

Intel 71

angry tapir writes "Intel is taking the first steps to implement thin fiber optics that will use lasers and light as a faster way to move data inside computers, replacing the older and slower electrical wiring technology found in most computers today. Intel's silicon photonics technology will be implemented at the motherboard and rack levels and use light to move data between storage, networking and computing resources. The new rack architecture with silicon photonics is a result of more than a decade of research in Intel's laboratories, Intel CTO Justin Rattner said. It could enable communication at speeds of 100Gbps and transfer data at high speeds while using less power than copper cables. The technology could also consolidate power supplies and fans in a data center, reducing component costs."

cancel ×

71 comments

Sorry! There are no comments related to the filter you selected.

Is this good-bye? (2)

c0lo (1497653) | about a year and a half ago | (#42614961)

Is this "good-bye OEM motherboards"?

Re:Is this good-bye? (0)

Anonymous Coward | about a year and a half ago | (#42615029)

Why would it be?

Re:Is this good-bye? (4, Insightful)

gl4ss (559668) | about a year and a half ago | (#42615035)

why? it's not like oem manufacturers do their chipsets today..

Re:Is this good-bye? (0)

zixxt (1547061) | about a year and a half ago | (#42615161)

Is this "good-bye OEM motherboards"?

Maybe.. Intel has done just as much damage to the hobbyist computer culture as Microsoft or Apple has maybe even more.

Re:Is this good-bye? (2, Interesting)

Pale Dot (2813911) | about a year and a half ago | (#42615345)

Is this "good-bye OEM motherboards"?

Maybe.. Intel has done just as much damage to the hobbyist computer culture as Microsoft or Apple has maybe even more.

Why? Intel has been very supportive of the most popular hobbyist OS, Linux!

Re:Is this good-bye? (5, Funny)

CajunArson (465943) | about a year and a half ago | (#42615459)

Oh yes, Intel's reign of terror that includes foisting tens of millions of systems that can easily boot practically any version of Linux and their insidious plot to use standardized system interconnects has truly ushered in an age of darkness from which the world will never recover. Don't even get me started on their insidious projects where they infiltrate the Linux kernel with completely open-sourced GPL'd graphics drivers! Truly they should all be put up on war crimes charges!

Now excuse me while I return to my secret resistance base where we are attempting to load updated ROMs on our Android phones. One of these weeks we'll download the right set of magic files & instructions from some random forum and hopefully not permanently brick the phone in the process. Only ARM can save us hobbyists from the tyranny of well documented and easily modified computing systems!

Re:Is this good-bye? (0)

Anonymous Coward | about a year and a half ago | (#42615853)

Careful! You don't want to get into the usual Slashdot "discussion" over what bricking really means

Re:Is this good-bye? (1)

sa666_666 (924613) | about a year and a half ago | (#42617179)

The sarcasm is strong with this one :)

Re:Is this good-bye? (0)

Anonymous Coward | about a year and a half ago | (#42617373)

I like the vacilating levels of irony involved.

Re:Is this good-bye? (0)

Anonymous Coward | about a year and a half ago | (#42618181)

Don't forget about Intel's evil conspiracy to bring closed source Steam games to Linux by helping implement better standards and better opensource drivers.

Re:Is this good-bye? (1)

blahplusplus (757119) | about a year and a half ago | (#42618615)

I know you're being funny but I think if you look at long term trends DRM has become more entrenched in the software world and small incremental changes on the hardware front and pressure from software industry/government to 'lock down' software, it could very well happen.

More and more new PC games lately have been MMO's, or a variations there-of. To think "It couldn't happen here" is naive at best given the attack on software and digital content ownership on multiple fronts.

Re:Is this good-bye? (0)

Anonymous Coward | about a year and a half ago | (#42624477)

DRMs have been in place since the beginning of time, don't you remember "don't copy that floppy" and all the obscure methods of locking software (ranging from finding the third letter on the fifth page of your manual to calling in to get the key). Hell, I'd say DRM was much worse 5 years ago when music was locked in tight (we now have DRM free music from services such as Amazon) and the horrible DRM attempts by game developers who required software installed whose specific purpose was DRM, thank goodness for sites such as GOG and to a lesser extent Steam (although steam itself is a minor form of DRM).

Re:Is this good-bye? (0)

Anonymous Coward | about a year and a half ago | (#42615489)

Aye.

Re:Is this good-bye? (1)

TheSkepticalOptimist (898384) | about a year and a half ago | (#42616217)

Need a new hobby other then building obsolete shoebox computers.

Re:Is this good-bye? (1)

Anonymous Coward | about a year and a half ago | (#42615311)

No no.. it's "HELLO to endless lifetime service contract with intel."

(captcha:bastards. wow thats weird. captcha can see the future)

Re:Is this good-bye? (0)

Anonymous Coward | about a year and a half ago | (#42617831)

Yep. This goes into production tomorrow, and OEMs are stopping production next week. Say "good-bye".

but (1)

Anonymous Coward | about a year and a half ago | (#42614995)

Don't electrical pulses along a copper wire go at the speed of light already?

I can see various advantages of course, but my meager knowledge of the subject suggests one of the main barriers is the encoding schemes for pushing bits about, and not the physical substance that the signals travel on per se.

Re:but (5, Informative)

Anonymous Coward | about a year and a half ago | (#42615019)

copper wires cause electromagnetic interference. Every single copper wire at those frequencies start to act as an antenna which causes problems. Light does not have this problem hence you can work with way much higher frequencies.

Re:but (0)

Anonymous Coward | about a year and a half ago | (#42615021)

Don't electrical pulses along a copper wire go at the speed of light already?

I can see various advantages of course, but my meager knowledge of the subject suggests one of the main barriers is the encoding schemes for pushing bits about, and not the physical substance that the signals travel on per se.

No. Pulses on a copper wire are slower than the speed of light.

Re:but (1)

Anonymous Coward | about a year and a half ago | (#42615047)

Speed of electromagnetic waves on copper are close to the speed of light (95 - 97%).
      http://en.wikipedia.org/wiki/Speed_of_electricity

Speed of light in optical fibre is approx 2/3 the speed of light
    http://en.wikipedia.org/wiki/Optical_fiber

Re:but (5, Informative)

NettiWelho (1147351) | about a year and a half ago | (#42615101)

Speed of electromagnetic waves on copper are close to the speed of light (95 - 97%). http://en.wikipedia.org/wiki/Speed_of_electricity [wikipedia.org]

Speed of light in optical fibre is approx 2/3 the speed of light http://en.wikipedia.org/wiki/Optical_fiber [wikipedia.org]

Propagation speed is affected by insulation, so that in an unshielded copper conductor ranges 95 to 97% that of the speed of light, while in a typical coaxial cable it is about 66% of the speed of light.[1]

Re:but (0)

Anonymous Coward | about a year and a half ago | (#42615249)

You're still talking about travelling fast enough to go around the earth around 4 times per second though.

Re:but (2)

ssam (2723487) | about a year and a half ago | (#42615411)

we are talk about the nanoseconds it takes a memory access requests to cross your motherboard and the answer to return. and the many CPU clock cycles spend idle while wait for it.

Re:but (0)

Anonymous Coward | about a year and a half ago | (#42615569)

You're still talking about travelling fast enough to go around the earth around 4 times per second though.

At 1GHz the electrons don't even have the time to travel from the power supply to the motherboard in a single cycle.
That is one of the reasons to why the PCB is littered with capacitors.
Reduce your CPU-speed to 1Hz and it won't be a problem.

Re:but (3, Informative)

Luuseens (1422579) | about a year and a half ago | (#42615783)

You'll find that electrons will need a whole lot of time to 'travel from power supply to the motherboard'. As per http://en.wikipedia.org/wiki/Speed_of_electricity [wikipedia.org] : "These speeds are on the order of millimeters per hour." It's the pulse (wave) that electron movement creates that is close to lightspeed, not the speed of electrons themselves.

Re:but (1)

rubycodez (864176) | about a year and a half ago | (#42617087)

the electrons don't really travel anywhere, it's alternating current after all. they just jiggle in an EM field. the velocity of propagation of changes in the EM field is the important thing.

Re:but (1)

Trogre (513942) | about a year and a half ago | (#42622177)

After going through the power supply it's all DC.

Re:but (1)

rubycodez (864176) | about a year and a half ago | (#42641771)

my computer has a clock line, and bus lines..

Re:but (1)

Trogre (513942) | about a year and a half ago | (#42709549)

That's true, however these are operating with pulsed DC, not AC as such. The clock signals oscillate from zero to a positive voltage, not a sine wave of positive and negative voltages, so the direction of the current does not change.

Re:but (1)

Anonymous Coward | about a year and a half ago | (#42615781)

You're still talking about travelling fast enough to go around the earth around 4 times per second though.

Why do people love to pull out the most convoluted statistics to help prove their point? Is the external hard drive next to my computer going to be connected by a cable that wraps around the earth twice? If not, then who gives a shit about around-the-earth performance? If you have a valid argument to make, then make it in a manner which actually fits to application being discussed.

Re:but (0)

Anonymous Coward | about a year and a half ago | (#42615453)

Hysteresis, my friend.... hysteresis. Speed means nothing without control.

wrong - blindly quoting textbooks out of context (2)

rubycodez (864176) | about a year and a half ago | (#42617025)

in waveguides the speed is slower, and that includes twisted pair and coax. Typical twisted pair impulse propagation is 45 - 65 % speed of light. Bare copper conductor over ground plane (what you quoted) is not relevant here.

Re:but (0)

Anonymous Coward | about a year and a half ago | (#42615111)

if fiber is 66% don't see how anything is gained on that front, velocity factor for the usual rg58 etc. is ~66%, but there are some types of coax that are +80%

Re:but (0)

Anonymous Coward | about a year and a half ago | (#42615211)

The main speed advantage will come from the data accuracy - you will no longer require a parity bit to validate the potential interference caused by electromagnetic interference

Re:but (1)

Anonymous Coward | about a year and a half ago | (#42623521)

Not so fast with the abandoning of error correction! Let's not go down that road.

In reality, there will be plenty of opportunities for data corruption.

Re:but (1)

Anonymous Coward | about a year and a half ago | (#42615381)

For the signal integrity illiterate: The propagation speed is about the same as copper wire normally is surrounded by a dielectric medium (y'know, plastics and such) glass can have similar properties. The whole advantage lies in the 1) signalling speed/bandwidth can go much higher on a given length of fiber, compared to the same length of copper wire or trace. And, of course, crosstalk/interference is non-existent using optics. Routing multiples of 10 Gbps lines on a motherboard across connectors is non-trivial due to all the electrical impedance changes, reflections and losses. This is much less of a problem in optics, allowing you to go much higher in bandwidth.

Re:but (4, Insightful)

dfghjk (711126) | about a year and a half ago | (#42615583)

AC steps in with a complete misunderstanding of the concepts.

Good job quoting Wikipedia though.

Re:but (4, Informative)

LordLimecat (1103839) | about a year and a half ago | (#42615977)

So are pulses of light in a fiber medium.

If you do the math, copper is actually around 10-25% lower transit latency. As has been pointed out, fiber wins because it suffers less interference and can go longer distances; that means fewer hops, higher frequencies, etc.

Re:but (1)

tilante (2547392) | about a year and a half ago | (#42616411)

Point of order: all electromagnetic waves move at the speed of light. It's simply that the speed of light in copper is slower than the speed of light in a vacuum.

Re:but (0)

Anonymous Coward | about a year and a half ago | (#42616617)

It's simply that the speed of light in copper is slower than the speed of light in a vacuum.

And can depend on the shape of the copper. And is frequency dependent. And depends on the properties of near by conductors and insulators. And in some cases, depends on the previous history of current and voltage having gone through the wire beforehand. Maybe referring to it as a a singular "speed of light" loses meaning when it depends on many other factors and that is why it typically gets referred to by the situational dependant and more generic "propagation velocity."

Re:but (1)

Anonymous Coward | about a year and a half ago | (#42617017)

>It's simply that the speed of light in copper is slower than the speed of light in a vacuum.

If you can actually get light to propagate in copper which is opaque in the first place... I am SURE that Bell Canada that advertise their VDSL as fibre would love to see that.

Re:but (1, Informative)

Anonymous Coward | about a year and a half ago | (#42615031)

It is not the speed but the bandwidth that is used. Light wavelengths don't cause interference like electronics do.

Re:but (2)

Jade_Wayfarer (1741180) | about a year and a half ago | (#42615053)

First of all, speed of light in one medium differs from speed of light in another. So yeah, probably there is some immeasurably small difference in speed of signal in copper and glass. But of course main difference comes from interference and heating - you can pack more channels of information running much higher frequencies in the same space. Plus there may be some speed gain from changing electronic elements with optronic ones (standard transistor can change its state only this fast - maybe optronic equivalent can do it faster?), but I don't have real numbers on my hands right now and too lazy to search and compare them.

Re:but (0)

Anonymous Coward | about a year and a half ago | (#42616661)

o yeah, probably there is some immeasurably small difference in speed of signal in copper and glass

The difference in propagation speed between different cables, fibers and over just air is easily measurable with a relativity low end oscilloscope and a room larger than a broom closet. That doesn't change the point that there are other reasons for switching to optical communication though.

Re:but (1)

Anonymous Coward | about a year and a half ago | (#42615077)

No, only light travels at the speed of light in a vacuum. EM signals in matter travel slower than light. In a typical coax, it's 60% of the speed of light in a vacuum ("c"). The bandwidth of fiber optics is higher. The real question is, is that data worth this effort?

Re:but (1)

dfghjk (711126) | about a year and a half ago | (#42621909)

Light IS an "EM signal". If light only travelled at the "speed of light in a vacuum" there would no need to qualify the speed of light with "in a vacuum".

AC does it again.

Re:but (1)

billyswong (1858858) | about a year and a half ago | (#42615395)

Pushing bits in a more packing way is of course important, but latency matters too. Optical circuits are more responsive.

Re:but (4, Informative)

smpoole7 (1467717) | about a year and a half ago | (#42615503)

> Don't electrical pulses along a copper wire go at the speed of light already?

That's not the problem, it's propagation effects and timing issues. As someone else here pointed out, these high-frequency signals are essentially radio waves and behave like radio waves. You have interference issues from other, nearby signals. The copper traces on your current motherboard must be carefully routed and kept at equal lengths (because they're essentially transmission lines), or you'll have some bits arriving later than others. Chaos. Using optical eliminates that problem.

(This is also why, if you've ever tried to repair a damaged motherboard, you probably weren't successful. Even if you could successfully identify all the damaged traces -- not easy, what with the "sandwich" layered design -- when you use little jumper wires to bridge the gaps, it just won't work reliably.)

By the way, these propagation effects are the reason why (counter intuitively) SATA and USB can more easily be made faster than older-style parallel connections. Once you get into the 100 megabit range, interference and the precise arrival time of the parallel bits becomes very hard to control. If it's a bit stream, even though it's several orders of magnitude faster, it's just easier to predict and control.

Re:but (1)

smpoole7 (1467717) | about a year and a half ago | (#42615533)

(A pedantic self-correction: it would be more accurate to say that the copper traces must be "controlled" lengths, rather than "equal." I was thinking "equal" in terms of a single, discrete data buss.)

Re:but (1)

petermgreen (876956) | about a year and a half ago | (#42616081)

By the way, these propagation effects are the reason why (counter intuitively) SATA and USB can more easily be made faster than older-style parallel connections. Once you get into the 100 megabit range, interference and the precise arrival time of the parallel bits becomes very hard to control. If it's a bit stream, even though it's several orders of magnitude faster, it's just easier to predict and control.

Then you have the likes of PCIe which make multiple serial links with indepdendent bit timing control then have a block of logic that synchronises and combines the bitstreams to make a higher bandwidth link.

Re:but (0)

Anonymous Coward | about a year and a half ago | (#42618679)

By the way, these propagation effects are the reason why (counter intuitively) SATA and USB can more easily be made faster than older-style parallel connections. Once you get into the 100 megabit range, interference and the precise arrival time of the parallel bits becomes very hard to control. If it's a bit stream, even though it's several orders of magnitude faster, it's just easier to predict and control.

Yes, you are right that the interconnect has gone serial. The high-speed parallel port of yore is now USB. The PATA hard drives are gone, leaving us with SATA. The PCI bus has gone PCIe, where every lane is a serial connection. In fact, the bitstreams are modulated and sent over these high-speed serial interconnects like across a modem connection.
In all these instances, the amount of DSP processing that happens in hardware to achieve those speed is incredible.

But the need for speed pushes manufacturers and to bundle multiple high speed serial interconnect together: for example PCIe x16, with 16 lanes. And we're back to a parallel bus. Except that this time around, instead of trying to push all the word bits simultaneouly through the bus, the bit stream is sliced into packets, and sent through a set of transmission lines like IP traffic would be sent through bonded telco lines.

In fact, the current and future bus protocols look like networking protocols more and more.

Re:but (0)

Anonymous Coward | about a year and a half ago | (#42618123)

No. Light doesn't even travel at the speed of light.

MTBF (1)

Anonymous Coward | about a year and a half ago | (#42615083)

Lasers and LEDs age. Does anyone knows whats would be the life of these "optical chips"?

Re:MTBF (1)

Viol8 (599362) | about a year and a half ago | (#42615173)

So do some normal electronic components, especially capacitors.

Re:MTBF (4, Funny)

MetalliQaZ (539913) | about a year and a half ago | (#42615481)

Just long enough so that you won't be upset about having to upgrade.

Endpoints (1)

gmuslera (3436) | about a year and a half ago | (#42615141)

The speed of the medium could be similar or faster to copper one, while there are no heat nor interference, but, what about the convertion between electrical and optical information? That could slow down things, induce heat, limit bandwidth or require more power? And there is that little factor called cost.

not new (5, Interesting)

thoper (838719) | about a year and a half ago | (#42615169)

relevant wikipedia:

http://en.wikipedia.org/wiki/Photonic_computing [wikipedia.org]
A claimed advantage of optics is that it can reduce power consumption, but an optical communication system will typically use more power over short distances than an electronic one. This is because the shot noise of an optical communication channel is greater than the thermal noise of an electrical channel which, from information theory, means that more signal power is required to achieve the same data capacity. However, over longer distances and at greater data rates, the loss in electrical lines is sufficiently large that optical communications will comparatively use a lower amount of power. As communication data rates rise, this distance becomes longer and so the prospect of using optics in computing systems becomes more practical.

and a more interesting article from 2010.
http://phys.org/news199470370.html [phys.org]

Today computer components are connected to each other using copper cables or traces on circuit boards. Due to the signal degradation that comes with using metals such as copper to transmit data, these cables have a limited maximum length. This limits the design of computers, forcing processors, memory and other components to be placed just inches from each other. Today's research achievement is another step toward replacing these connections with extremely thin and light optical fibers that can transfer much more data over far longer distances, radically changing the way computers of the future are designed and altering the way the datacenter of tomorrow is architected.

Marketing Guff / Fibre Channel (0)

Anonymous Coward | about a year and a half ago | (#42615291)

Can somebody cut through the marketing stuff and explain the difference between this and Fibre Channel please?

Re:Marketing Guff / Fibre Channel (0)

Anonymous Coward | about a year and a half ago | (#42615531)

Can somebody cut through the marketing stuff and explain the difference between this and Fibre Channel please?

Of course not. Intel has people paid to spread the news that they just discovered the wheel.

Re:Marketing Guff / Fibre Channel (0)

Anonymous Coward | about a year and a half ago | (#42615745)

Smaller, less power, faster, maybe even cheaper?

Déjà vu (1)

vmlemon (1203598) | about a year and a half ago | (#42615519)

I'm sure that Intel called this "Light Peak" initially, when they demonstrated it two years ago.

Re:Déjà vu (0)

f()rK()_Bomb (612162) | about a year and a half ago | (#42618165)

Light Peak is the codename for the Thunderbolt interface, you know the one on the iphone 5 and such.

Only 100 Gbps? (0)

Anonymous Coward | about a year and a half ago | (#42615625)

Seems to me that this technology could achieve transfer speeds a lot greater than 100 Gbps.

repeat from 1989 (0)

Anonymous Coward | about a year and a half ago | (#42615755)

Ever since fiber optics was invented we've been hearing about how optical computing and optical data transfer is going to revolutionize computing. It's not that I don't believe that it won't happen some day, but you've got to invent a mirrorless, efficient optical switch first. Otherwise you're never escaping the "slow" copper wire at some point.

Uh... (0)

Anonymous Coward | about a year and a half ago | (#42616301)

hasn't IBM been doing this for more than a decade?

No way (1)

ilsaloving (1534307) | about a year and a half ago | (#42616613)

If there arn't LEDs spread throughout the interconnect that flash randomly during operation, then how am I supposed to accept that this is the future in server technology?

Decades my ass (0)

Anonymous Coward | about a year and a half ago | (#42617889)

Fiber optics were among the first technologies exchanged for human biological material.

I don't understand the physics of the statement. (0)

Anonymous Coward | about a year and a half ago | (#42618527)

From my understanding, light travels 2/3rd the speed it does in a straight copper wire. It's pretty simple to expand bandwidth via providing more parallel links of communication, much cheaper than via separate laser/detector modules. So what's the point if it's slower than direct wires?

SO 15 seconds ago.... (1)

Crypto Gnome (651401) | about a year and a half ago | (#42620229)

100Gbps is inadequate today for a core system interconnect.

You can buy 100Gbps NETWORK interfaces (although, obviously not for your desktop PC, because, as per above THE BUS IS TOO SLOW)

I'll say it again, releasing technology to deliver 100Gbps interconnects at the system core is holding back innovation TODAY, not "in the future" - TODAY FOLKS!

Re:SO 15 seconds ago.... (1)

timeOday (582209) | about a year and a half ago | (#42620507)

Getting technology that was only available in a $60,000 router into my PC at an affordable price is what I'd call progress. The article says Intel had optical Thunderbolt at CES. I would be delighted if I could run a $50, 50m fiber to extend consoles to different parts of my house instead of having 3 PCs running 24/7. (Of course PCs themselves are getting so cheap and power efficient that the race is still on...)

Further coverage from 3 months ago. (0)

Anonymous Coward | about a year and a half ago | (#42622091)

A very similar tech from IBM was covered by Slashdot last month slashdot.org/story/12/12/10/1433220/ that was promising to "scale to peta- and exabit speeds".

Check for New Comments
Slashdot Login

Need an Account?

Forgot your password?