intel introduces lasers from standard silicon

[kcthomas]

Text

so will we see fiber optics routing signals around the chip? might this be a solution to long wires on a chip?

dont know much about fiber optics, is there any transmission line effects similar to long wires on a die? will we be able to send a optic signal and pick it up on the other side with little to no signal degradation and no repeater like circuitry?

 

[walla]

Though i'm no expert, I definitely think that optical signal transmission...even optical logic gates...are the future of computing.

The main problems with signals propagating along long wires on the die are IR drop, capacitance, which often necessitate repeating stages as you've mentioned. As die sizes decrease, wires become more narrow and are placed closer together, amplifying the problem.

With optical signal communication, there is no IR drop or capacitance to speak of. The propagation speed/signal fidelity of optical signals is largely a function of the refractive qualities of the core/cladding and the "straightness" of the waveguide.

The attenuation of optical signals are primarily due to scattering (photons changing direction due to hitting certain dieletric particles), absorption (photons converted to other forms of energy such as heat) and bending loss (photons escape the waveguide due to bending of the waveguide). The main problem with using optical signals in modern computing is that wires need to bend at extremely sharp angles, which allows most of the photic energy to either reflect backwards or escape the waveguide itself. Traditional waveguides prove impractical for this purpose.

However there have been some advancements in photonics regarding photonic crystals and waveguides. These have a different structure than traditional, and can contain much more photonic energy around such sharp bends. Without going into specifics...this property could allow optical routing in computer nanoarchitecture. But I feel this is quite a ways down the road.

 

[Fencer128]

Hi,

The lack of an acceptable optical "transistor" is going to hamper optical processing for a while yet. The big news about "Si lasers" (apart from the fact that they've managed to get Si CW lasing at all) is that:

a) This could theoretically be a good step forwards with regard to optical chip-chip interconnects. Si lasers on Si detectors would enable very compact and more easily grown structures. As a previous poster said, photonic crystal structures could perhaps be exploited in this regard, though I imagine that free-space interconnects are perhaps currently the most advanced technology in this area (much less losses)

b) Any application that is more power than wavelength dependant and that currently uses GaAs, GaNi, etc. lasers in conjunction with logic on Si can now produce an integrated package. Again, the scale and growth costs should hopefully come down because of this.