backtop


Print 60 comment(s) - last by EricMartello.. on Feb 11 at 9:00 AM


A new type of germanium laser may bring about the fastest room temperature computer communications yet.  (Source: Christine Daniloff)

Germanium is a versatile metalloid that is found in zinc-copper-lead ore bodies and coal. Germanium is used to create special semiconductors and tunable terahertz lasers.  (Source: Wikimedia Commons)
Lasers could be used for faster bus between CPU and memory

Recent advances in processor technology have lead to a bottleneck in communication with memory. Trying to overcome this barrier with the currently used method -- smaller, better designed semiconductors -- is impractical due to the higher power necessary to transport data at faster rates. Scientists at MIT believe they may have discovered a way to circumvent this limit -- use germanium lasers to transport data at the speed of light through air.

Researchers at MIT have demonstrated the first germanium laser functional at room temperature that can produce wavelengths of light useful for optical communication. The scientists hope to apply this new technology to the world of processors and begin moving data and (possibly) performing calculations using light instead of electricity.

The process of integrating these new components into the chip has the potential to be very expensive, however. Assembly of a chip is an extremely difficult process in which materials are layered onto a silicon wafer and patterns are etched into them. Additionally, materials must chemically bonded with layers above and below it in a chemical and thermal environment non-caustic to all the rest of the materials used during assembly.

Traditionally used lasers like gallium arsenide serve as an example of the difficulty of incorporating un-favorable materials; in addition to the fact that gallium arsenide is more expensive than silicon to begin with, the lasers have to be grafted to the semiconductor separately. Fortunately, germanium may avoid some of these integration problems as it is already being integrated into manufacture processes for its ability to increase chip speed.

From more of a pure research perspective, this work proves an important hypothesis: indirect-band-gap semiconductors can in fact produce practical lasers. This fact comes as a counterpoint to the generally accepted opinion that indirect-band-gap semiconductors will never produce laser light. The formerly accepted opinion comes from the tendency for these types of materials to produce heat rather than light, a tendency that MIT scientists overcame through such methods as doping (adding atoms from another element).

The director at Massachusetts-based Analog Devices Semiconductor offers excellent insight into this new technology, explaining that, “High-speed optical circuits like germanium in general, that’s a good marriage and a good combination. So their laser research is very, very promising.”

Miao continues, pointing out that the germanium lasers need to become more energy-efficient before they can truly be considered a practical source of light for optical communications systems. “But on the other hand,” he says, “the promise is exciting, and the fact that they got germanium to lase at all is very exciting.”

Further reading on the engineering required to accomplish this feat can be found in this letter, which is to be published in the journal Optical Letters.




Comments     Threshold


This article is over a month old, voting and posting comments is disabled

Germanium?
By MegaHustler on 2/5/2010 3:58:40 PM , Rating: 5
Germanium? I suspect it will be well-built, but overpriced. Wait for the Japanium model...




RE: Germanium?
By MadMan007 on 2/5/2010 5:42:01 PM , Rating: 5
mh, either way you'll have to be aligned properly with the Axis to get it to work.


RE: Germanium?
By erple2 on 2/5/2010 6:00:30 PM , Rating: 5
No. I'd expect it to be significantly over-engineered, and consequently over-priced and not as reliable as it should be.


RE: Germanium?
By daar on 2/5/2010 6:41:18 PM , Rating: 5
Yes, the Japanium model, which might seemingly be competitive in terms of price/performance ratio, until people realize it is actually quite inferior and lead to many crashes and/or explosions. True value seekers will have to turn to...


RE: Germanium?
By TSS on 2/6/2010 8:51:22 AM , Rating: 2
Chinesium?

Though that'll likely be a cheap inferior copy of Germanium so it might be best to just stick with the original.


RE: Germanium?
By saiga6360 on 2/6/2010 11:21:46 PM , Rating: 5
Well, there is always Americanium. It has the advantage of becoming too big to fail.


RE: Germanium?
By Diesel Donkey on 2/7/2010 7:30:24 PM , Rating: 2
You mean Americium?


RE: Germanium?
By overzealot on 2/8/2010 3:08:24 AM , Rating: 3
I'd say Americum, but without an I there's no profIt.


RE: Germanium?
By RU482 on 2/8/2010 11:23:19 AM , Rating: 2
that would take one large tissue


RE: Germanium?
By descendency on 2/6/2010 4:44:23 AM , Rating: 2
Then in five years the Chinese immitation model will be widely available. Better known as KIRF.


RE: Germanium?
By thekdub on 2/6/2010 12:51:42 PM , Rating: 1
I'd wait until after the recalls on the Japanium models. I hear that the laser has a tendency to stick and speed up the processor, causing the system to crash.


RE: Germanium?
By chrnochime on 2/7/10, Rating: 0
RE: Germanium?
By neogrin on 2/8/2010 11:06:35 AM , Rating: 1
You have learned how to Troll.

Wow good for you.


Speed increase
By freeagle on 2/5/2010 1:37:00 PM , Rating: 2
In theory, how much faster can the trasfers be when light is used instead of electricity?




RE: Speed increase
By neo64 on 2/5/2010 2:05:40 PM , Rating: 1
signalling using copper wires is established by electric fields in the wire between the devices.
the individual electrons (in the wires) per se move slowly, in the order of metres/sec; but the electric field is established at light-speed..

now since light is also made up of electric fields and is, you know, light; it travels at light-speed as well

take 10Gbps ethernet (fibre optic cable) vs 1Gbps ethernet (copper cable) as an example
the advantage in these optical systems is the massive bandwidth available.


RE: Speed increase
By Yawgm0th on 2/5/2010 4:12:07 PM , Rating: 2
Alright amateur DT physicists and computer scientists...

This is accurate in that electrons travel very slowly, but that has absolutely nothing to do with why fiber optic is better (not "faster") than copper.

For starters, let's fix your example. Ethernet at 10gbps or 1gbps is Ethernet. Ethernet operates on the data link layer over any conceivable medium. Specifically, 10GbE can operate over copper or fiber, as can GbE. It's ridiculous to assert that fiber optic has inherently more "bandwidth" because it moves faster than electrons.

Light, however, is not affected by EMI or electrical signal attenuation, conceivably allowing for more more data to be transmitted over either longer distances or in smaller spaces.

There are physical limits to how quickly electrical signals can be transmitted, and we are approaching or have reached them already in some specific applications. Light is not susceptible to the same limitations.

That's why using some form of light to transmit data is superior. It has very little to do with how quickly light travels.


RE: Speed increase
By Flunk on 2/5/2010 4:54:44 PM , Rating: 2
Don't the big friction issue. Computers that don't generate heat by friction would be very cool running.


RE: Speed increase
By boogle on 2/6/2010 5:13:54 PM , Rating: 2
Just as an extra to this:

Fibre optics can theoretically transmit a lot more information by using multiple wavelengths. So an existing fibre optic cable could be 'upgraded' by using more sensitive equipment at either end of the cable. By using ever smaller 'chunks of colour' you can increase your effective bandwidth.


RE: Speed increase
By neo64 on 2/7/2010 5:59:36 AM , Rating: 2
i merely pointed out to the OP that speed in both mediums is a non-issue (which was his original question) and that the real difference was something else eg bandwidth advantage
quote:
This is accurate in that electrons travel very slowly, but that has absolutely nothing to do with why fiber optic is better (not "faster") than copper.

nowhere in my post did i imply that the electron drift velocity somehow limits wired transmission


RE: Speed increase
By delphinus100 on 2/7/2010 3:47:26 PM , Rating: 2
Would there not also be less heat dissipation for the same performance?


RE: Speed increase
By mcnabney on 2/8/2010 10:46:32 AM , Rating: 2
That is the problem. Using light itself doesn't generate heat, but the researchers in the article did comment that the amount of heat being generated now would make it unusable.

Personally, I don't see the point. It is useful in real fibre optics to spend the time/energy changing the signal states from electrons to photons - we are talking about large distances and maximization of speed/capacity. These guys are talking about using an optical bus, which would eliminate the delay between subsystems/controllers in a computer. But these scientists haven't been paying too much attention to computers these days. Everything is moving onto the chip. Memory controllers, bridges, IO, and graphics. I'm just not sure there will be that many workable applications even if they eliminate the heat issue. Also, how much time will the state-change to photons and back again consume? It has to be practically nill for there to be benefits to this when the targeted distance is a few inches.
And capacity really shouldn't be an issue. Need more capacity, trace more wires


RE: Speed increase
By mmcdonalataocdotgov on 2/8/2010 6:37:26 PM , Rating: 2
To add, at the distances they are talking about here, there must be some appreciable time delay to change from electrons to photons and back again at the other end of the bus. Of course, on a small die, you could mux/demux a lot wider memory bus with this technology.

Anyway, I still think optics are for distance and bandwidth, not on-die maneuvers like this. Let's have an article on some phase change breakthrough, and then the quarterly DT optics bus breakthrough post will mean something.


RE: Speed increase
By EricMartello on 2/11/2010 9:00:53 AM , Rating: 2
quote:
Light, however, is not affected by EMI or electrical signal attenuation, conceivably allowing for more more data to be transmitted over either longer distances or in smaller spaces.


Right, but there is still an attenuation factor that occurs as the light passes through the optic fibers themselves.

Another potential issue with "optical circuitry" is that if it isn't 100% optical and is a hybrid of electronic and optical circuitry, there will need to be a bridge at some point that either translates the optic signals into electric, or combines them. This process would undoubtedly be a bottleneck and may limit the real-world benefits of an "optical" processor.

But the nice thing is that heat is not going to play as big of a role in this type of system as it does now with our electronic circuits.


RE: Speed increase
By geddarkstorm on 2/5/2010 2:08:34 PM , Rating: 2
Not really /faster/ in sheer speed per se, that isn't the point. The point is light has an enormously larger bandwidth. You can layer it, with multiple signals happening at the exact same time by putting them on differing wavelengths. So, while with electricity, you can only send one bit of data per cycle (hertz) per path, with light you can send dozens, nigh limitless in theory, of bits per cycle per path depending only on the wavelength spread of the emitter and sensitivity of the detector.

Add to this the ability of light to get quantum entangled, allowing light itself to perform calculations while in flight... And you can see how vastly superior to pure electronics a light system is.


RE: Speed increase
By MrPickins on 2/5/2010 3:07:08 PM , Rating: 5
It also eliminates crosstalk across the bus which electrical signals are prone to.


RE: Speed increase
By greylica on 2/5/10, Rating: 0
RE: Speed increase
By wifiwolf on 2/8/2010 7:45:01 AM , Rating: 2
Yes it does, but in the long term - when using multple frequencies we need to take care about interference which could be comparable to crosstalk.


RE: Speed increase
By Yames on 2/5/2010 5:44:01 PM , Rating: 3
You can put "multiple signals" or broadband on copper.


Star Trek?
By superPC on 2/5/2010 3:15:51 PM , Rating: 2
Optronics and isolinear chips really can be our computational devices in the future.




RE: Star Trek?
By EJ257 on 2/5/2010 4:08:28 PM , Rating: 2
Not too far from the truth. IBM is already working on a 3-D optical storage cube thing although I have a feeling the final product will look more like a Jedi holocron than a Starfleet isolinear chip. This thing is another piece of the puzzle. Now all we need it a static warp field and we'll have FTL computing. :)


RE: Star Trek?
By MasterBlaster7 on 2/5/2010 8:53:07 PM , Rating: 2
Actually...I think quantum computing is faster than light computing...if you consider the entangled element...unfortunately we cannot read that data faster than light...all communication is classical in quantum transmissions.


RE: Star Trek?
By jimhsu on 2/8/2010 2:56:16 AM , Rating: 2
Yes, an important point. In order for quantum entanglement to be of any use for communication, you need a classical channel with which to send two bits for every quantum bit. Hence, it is useless.


RE: Star Trek?
By Xavi3n on 2/5/2010 4:12:10 PM , Rating: 3
This, it really amazes me how forward thinking star trek (and by extension The Next Generation) was.


RE: Star Trek?
By descendency on 2/6/2010 4:49:50 AM , Rating: 2
It is sort of a self fulfilling prophecy. Many of the "nerds" working on this stuff now were likely science fiction geeks in there youth. So there minds are ultimately aiming at creating those worlds even if it is subconscious. I call it self fulfilling because their imagination was used to create much of the tech when they were reading books and stuff.


RE: Star Trek?
By porkpie on 2/7/2010 2:02:54 AM , Rating: 3
I hope you were kidding. Other than a few of the TNG episodes actually written by real SF authors, there's almost no science in Star Trek's SF...and pretty much anything postulated in the series that is actually well-grounded are things that were first introduced and beat to death in a thousand previous SF stories/books/films.


Wait for the Unobtainium version
By Lord 666 on 2/5/2010 1:31:56 PM , Rating: 5
Now explains why Unobtainium costs $20 million a kilo.




By geddarkstorm on 2/5/2010 2:51:32 PM , Rating: 5
I'm assuming due to inflation, by that time $20 mil is the average price of a happy meal.


RE: Wait for the Unobtainium version
By greylica on 2/5/2010 8:47:42 PM , Rating: 2
That's why they are going to Pandora's planet...


By StevoLincolnite on 2/6/2010 12:56:27 PM , Rating: 2
...To meet the Ewoks? And drink cups of tea?


IBM did this a few years ago, maybe mid 90s?
By iFX on 2/5/2010 1:46:24 PM , Rating: 2
Hard to remember. The project worked but was prohibitively expensive and the actual unit was huge. Most of the kewl "new" ideas floating around were thought up my IBM decades ago, the technology just wasn't there or it was too expensive at the time.




RE: IBM did this a few years ago, maybe mid 90s?
By iFX on 2/5/2010 1:50:34 PM , Rating: 1
Typos abound.


RE: IBM did this a few years ago, maybe mid 90s?
By ZmaxDP on 2/5/2010 8:43:44 PM , Rating: 2
Wow, did you just get downrated for picking on your own post??? That's funny...


By iFX on 2/5/2010 11:27:56 PM , Rating: 2
It looks that way! LOL


By messyunkempt on 2/5/2010 3:46:59 PM , Rating: 5
quote:
the actual unit was huge


Must.... resist..... the innuendo....


By HrilL on 2/5/2010 4:51:54 PM , Rating: 2
I remember reading about IBM making an optical CMOS Chip around 2006 I believe.


correction
By tensor9 on 2/5/2010 2:50:31 PM , Rating: 3
One small correction: germanium is not a transition metal. It's quite firmly a main group element.




RE: correction
By zozzlhandler on 2/5/2010 4:51:51 PM , Rating: 1
Absolutely correct. I suspect the writer was thinking of "Group IV" elements (C, S, Ge,...) and used the wrong (*very* wrong) word. Group IV are the semiconductors, halfway between metals and non-metals.

DT, would you like to hire me as an editor?


RE: correction
By zozzlhandler on 2/5/2010 4:53:01 PM , Rating: 4
Oops, I mean Si, not S.

Aaargghh.


RE: correction
By HotFoot on 2/5/2010 5:27:04 PM , Rating: 5
Hehe, irony.

I kind of think of every poster here as a bit of an editor... or at least, a lot of the time corrections/information from the posts made make their way into the article.


Lightspeed interconnects?
By Motoman on 2/5/2010 1:26:06 PM , Rating: 4
Bah. Wake me up when we can get past mere lightspeed. I expect at least Ridiculous speed, if not Ludicrous speed.




RE: Lightspeed interconnects?
By johnsonx on 2/5/2010 1:28:34 PM , Rating: 2
lol, careful, don't go plaid.


RE: Lightspeed interconnects?
By HotFoot on 2/5/2010 5:28:45 PM , Rating: 3
This is me, giving you the raspberry.


RE: Lightspeed interconnects?
By thekdub on 2/6/2010 12:55:28 PM , Rating: 2
But nobody has ever gone that fast before! Shouldn't you buckle up first?


:-)
By Oregonian2 on 2/5/2010 1:49:07 PM , Rating: 2
Back to the future, the return of Germanium!

I haven't seen an ad from what's his name who did Germanium ads for the last remaining germanium semiconductor company (that faded into history some time ago). Ollie Germanium was it? :-)

P.S. - For those who haven't been around for a while, semiconductors were germanium based prior to the newfangled silykon coming into fashion. :-)




RE: :-)
By porkpie on 2/5/2010 2:00:47 PM , Rating: 2
You're right about the rest, but high-speed germanium switching transistors have long been (and still are) heavily used in RF and other related fields.


materials fight!
By MadMan007 on 2/5/2010 5:43:47 PM , Rating: 4
Future materials fight gameshow season 1:

Graphene versus Germanium...fight!




Excellent
By amanojaku on 2/5/2010 1:34:55 PM , Rating: 2
We are one step closer to my shark-mounted, laser-powered PC. The water for the sharks will also provide cooling. Genius. Pure mad genius.




By rmlarsen on 2/5/2010 4:39:49 PM , Rating: 2
If by recent you mean with the advent of the von Neuman computer architecure in the '40s then yes, I agree ;-) But faster memory interconnects will certainly allow the likes of Intel to put more cores in a single package without starving the excution units.

While the technology changes over time, it will always make economic sense to have a fairly deep memory hierarchy.




By nofumble62 on 2/6/2010 1:53:01 AM , Rating: 2
Still haven't seen that yet except, maybe in few high end IBM system.




LOL
By blppt on 2/6/2010 1:41:19 PM , Rating: 1
I knew...even before I clicked the link to this story that there would be a picture of Dr. Evil on top.

+1000 to Dailytech. ROTFL.




"If you can find a PS3 anywhere in North America that's been on shelves for more than five minutes, I'll give you 1,200 bucks for it." -- SCEA President Jack Tretton
Related Articles
The Future is Indeed Light
August 1, 2008, 11:04 AM













botimage
Copyright 2015 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki