IBM chip technology delays speed of light

Nathan Davis
28 December 2006, 3:29 AM


Send to a friend Print

IBM says it can greatly increase computer performance by slowing the flow of light with a ground-breaking device the size of one square millimetre.

IBM researchers have unveiled a device that can delay the flow of light on a silicon chip. In doing so, they believe this is a great step closer to greatly increasing the performance of computers by internally using optical communications as opposed to electrical.

Light can carry more information, and faster, than that of electrical signals. Because of this, these researchers believe that using light signals for transfering data around within computer chips will result in a drastic performance increase compared to what we currently have available.

"As more and more data is capable of being processed on a chip, we believe optical communications is the way to eliminate these bottlenecks. As a result, the focus in high-performance computing is shifting from improvements in computation to those in communication within the system," says Dr. T.C. Chen, vice president of Science and Technology for IBM Research.

"Today's more powerful microprocessors are capable of performing much more work if we can only find a way to increase the flow of information within a computer."

So why delay the flow of these signals? Delaying, or rather, buffering a data signal by temporarily holding it on a chip, whether electrical or optical, is necessary in order to control the flow of information. This is a crucial aspect of reliable data transmission.

The impressive advancement with this device is not the ability to slow the flow of light, but rather the size of the device that can do the work.

In the past, light was simply passed through optical fibers in order to delay it, however these light buffers were far too large to include on today's ever-shrinking microchips (we assume, say, the size of your Christmas tree).

According to IBM, in order to practically include such buffers, or 'delay lines' on microchips, they must be smaller than one square millimetre. IBM scientists have managed to do just that.

"This advancement could potentially lead to integrating hundreds of these devices on one computer chip, an important step towards on-chip optical communications."


Post your comment



Comments

 RSS feed Email alert

max rubin:

No matter what IBM still commands vast amounts of technological power and innovation as well as the capital and marketing.

29 February 2008, 8:29 PM (3 years ago)report abuse Send to a friend reply

Bill Hermonat:

I'm surprised that we are considering that future optical chips will be faster because of how fast light travels. Maybe we have forgotten that electricity ALSO travels at the speed of light.

Where we will most likely see gains in the manufacturing of computer processors is MORE information traveling at once (i.e. parallel processing) resulting in faster computations.

We may also see a gain in lower heat build-up within processors.

Another benefit we will achieve will be in advances of optical storage, as mediums become more 3-dimensional in storage, rather than 2-dimensional (storage on surfaces) as they are more commonly used now.

In any case, non of the gains we see will be because of the speed of light. That is old news in using our current technology.

29 February 2008, 8:35 PM (3 years ago)report abuse Send to a friend reply

Steve harker:

I assume it will be more in chip interconnects well at least initially as a method to ship data at very high bandwidths from one chip to another or from one functional block to another on die of same chip, just as fiber is used to overcome limitations of copper, on chip optical connections would remove RC delay problems and allow throughput to increacse without the propogation delay problems associated with metal layers and low K dielectrics. And I assume some small light delays may be required to keep various data streams in sync.

29 February 2008, 8:35 PM (3 years ago)report abuse Send to a friend reply

Hector:

Bill, correct me if I'm wrong, but electricity and current are possible because electrons are shared at the atomic level in metals and other elements who have electrons on outer shells and can easily share those among atoms, thus the speed of the current is bounded to the material used to transmit it, the easier the electrons are shared the less resistance and thus less heat the whole process generates,

So by using light, are no longer limited by matter, you will be using photons, and light-sensitive receptors which will bring advantages in speed, heat, power compsumption, etc.

Correct me if I’m wrong...


29 February 2008, 8:35 PM (3 years ago)report abuse Send to a friend reply

Bill Hermonat:

Short answer on this is .. "no".

In response to your comment:

"So by using light, are no longer limited by matter, you will be using photons, and light-sensitive receptors which will bring advantages in speed, heat, power compsumption ..."

I would have to look this up, but I don't recall any readings on the speed of electricity, which travels at the speed of light, being affected by the mediums they travel in, whether that be copper, steel, water ... or any other source. In most cases this is true for light as much as it is for electricity. There are some exceptions for light that I am aware of, but only under very extreme and controlled circumstances usually conducted in experimentation.

Light, like electricity, also needs a medium to travel through. In this case we are talking about fiber-optic cable. This can be made of glass, plastic or other synthetics. So if electricity can be slower than the speed of light because of the medium, so can light be reduced because of the medium.

However, I believe you will find that most common mediums deliver both at the same speeds without any measurable differences.

Again the only advantages I can believe we might achieve by using light in computer chips are in lower heat. I don't know if this might reduce power consumption. That's just speculation at this point. Interconnect might be faster as well, as someone else pointed out.

Thanks for your interest and comments.

Bill

29 February 2008, 8:35 PM (3 years ago)report abuse Send to a friend reply

rdodemotta:

The electricity travels at differents speeds.

In a CTR, travels almost at the same speed of light, but only almost

In a battery, is no moving at all!

And a light, no need a medium to travel. It can travel in space!. Light is an electromagnetic wave

29 February 2008, 8:35 PM (3 years ago)report abuse Send to a friend reply

Ryan:

You are absolutely correct about the parallel processing being the main boost in computing power in optical chips versus electrical chips. The speed of light in a fiber is actually a little bit faster than the speed of current on a wire and can be faster or slower depending on the metal used, the gauge of the wire, how much current is flowing and the temperature. Light on a fiber does not have any of these issues. Transmission of the light is not much of a problem as it doesn't degrade nearly as quickly based on distance. The main issues are producing and reading the light on either end since it's not just an on/off bit, but many, many bits transmitted at the same time on different wavelengths.

29 February 2008, 8:35 PM (3 years ago)report abuse Send to a friend reply

Bill Hermonat:

You say:

"The speed of light in a fiber is actually a little bit faster than the speed of current on a wire and can be faster or slower depending on the metal used, the gauge of the wire, how much current is flowing and the temperature. Light on a fiber does not have any of these issues."

I don't believe this to be accurate, that the speed of electricity is affected by the medium and that it travels slower than light. And that there is any measurable difference in the speed of electricity from one medium to another, unless you are talking about a laboratory-controlled experiment that you would never see in commercial use.

If you know this to be true, please provide a source for your findings.

Thanks!

Bill

29 February 2008, 8:35 PM (3 years ago)report abuse Send to a friend reply

Steven:

If you trust Wikipedia, and some don't, it sounds like they are right. The speed of an electric current is affected by the medium. But as you will see if you read the article, its not always that easy nor that simple.

http://en.wikipedia.org/wiki/Current_(electricity)

29 February 2008, 8:35 PM (3 years ago)report abuse Send to a friend reply

Anonymous:

Light is instantaneous especially at such a short range ie in a computer. However copper transmission is not instantaneous.

It is true that through different mediums the flow of electricity is different. This is because of free electrons which enable electricity to be transmitted. If a substance such as steel were to be used it would greatly affect the time in which electricity is transmitted because steel has few free electrons. Whereas substances such as gold and copper have many free elctrons which enables the electricity to flow at a higher speed.

When we are talking microseconds in a computer for data transfer that delay the steel would cause will affect a great deal.

Even with a copper medium however there is a delay of transmission. The copper needs to be able to build up its many free electrons so that it can flow through the copper each time electricity is applied to it. Whenever the copper is in a rest state ie no electricity is applied, the free electrons move making the transmission of electricity delayed.

Ryan McAuliffe

29 February 2008, 8:29 PM (3 years ago)report abuse Send to a friend reply

Anonymous:

This is a really sad discussion, did no one do physics at school? Light and electrons cannot go the same speed. Why? Electrons have mass where light does not. Electrons can get accelerated to very high speeds but it takes effort and requires a good medium (like a vacum).

Which brings me to the other question, yes the medium matters thats why you have a vaccum in your cheap (ie non LCD etc ) TV set so the electrons can flow relatviely freely.

29 February 2008, 8:35 PM (3 years ago)report abuse Send to a friend reply

Tony:

I haven't read all the messages, but from what I see you are all missing a very important point. Information is not transmitted through a wire by electrons traveling from one end to the other. That would simply be a DC current. Information is transmitted as waves. Same as sound travels trough the air. It's not air traveling from the sound source to your ears; it's just an oscillation. Imagine wind blowing at the speed of sound! Electron speed is low but the speed of the signal is very high. It is in fact faster than light through optic fiber. Information through optic fiber is transmitted by photons traveling from one end to the other, and it is slower. So why is optic fiber better? It's all about information DENSITY. Optic fibers can carry a lot more information at once. And there's also a matter of signal quality, EMF interference, security, etc.

29 February 2008, 8:35 PM (3 years ago)report abuse Send to a friend reply

Anonymous1:

ye... Tony is right

29 February 2008, 8:38 PM (3 years ago)report abuse Send to a friend reply

vsXver:

roll

29 February 2008, 8:38 PM (3 years ago)report abuse Send to a friend reply

Not signed in
anonymous user Anonymous user