Sun Wants To Replace Wire With Laser

H

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
According to the New York Times, Sun Microsystems has a contract with the government to develop a new generation of processors that use lasers instead of wires. Obviously the benefits of using lasers over traditional wires would be huge…if they can get it to work.

The technology, part of a field of computer science known as silicon photonics, would eradicate the most daunting bottleneck facing today’s supercomputer designers: moving information rapidly to solve problems that require hundreds or thousands of processors.
Click to expand...
 
quite a daunting task I think. Laser requires precision and I'm thinking 24 million lasers reaplacing switches in processors would take a crapload of calibration. Jostle that box of CPU and it's dead,

Imagine the lasers crossing the streams!

Is that bad? ;)
 
The Sun researchers acknowledge that their project is a significant gamble.

“This is a high-risk program,” said Ron Ho, a researcher at Sun Laboratories who is one of the leaders of the effort. “We expect a 50 percent chance of failure, but if we win we can have as much as a thousand times increase in performance.”
Click to expand...

... a "gamble" and "high-risk" that Sun isn't even paying for...

Neat though. Cool to see this is coming around. I feel a quote coming on from a certain someone....
 
This is the same company (Sun) that switched to AMD CPU's because their own house tech (Sparc) couldn't keep up with the competition. Now the US is going to give that same company $44 million of tax payer money to fund a 50/50 shot at success?

I don't see the rationale. How does it make sense to spend $44 million because some underperforming tech companies claims it might be able to produce chips 1000% faster? Send me $4 million and I might be able to make chips that are a trillion times faster.
 
The probability of you doing so is very near zero so an investment of 44 million on a 50/50 return is a far better gamble than giving you 4 million :p
 
i dont understand the reasoning, dosent electricity travel at the speed of light?

unless your talking about ramp up/down time for the signal on the line, but even then lasers have a ramp up/down time as do the opto-sensors...

this seems like sun just made something up and asked for $44 million dollars...
 
This is weird, because I was just reading an article in my Wired magazine about singularity, and it was talking about how people always see technology and things as growing linearly, but a lot of people think it will actually be exponentially, so we're eventually going to have huge leaps in technology, and all these things people see as not happening until far far into the future will actually happen sooner than we think. So maybe this is one of those huge leaps. And then before we know it, the robots are going to become aware...



or maybe not, who knows?
 
First they need to make Java turn into an actual coffee maker and then I will believe their sorcery.
 
IBM/Motorola was working with switching light instead of electrons several years ago. They claimed they'd have it to market by ~2010 if I remember right... like so many other things.
 
Azhar said:
quite a daunting task I think. Laser requires precision and I'm thinking 24 million lasers reaplacing switches in processors would take a crapload of calibration. Jostle that box of CPU and it's dead,

Imagine the lasers crossing the streams!

Is that bad? ;)
Click to expand...
Refer to the source article...

I believe they are planning on using 'lasers' as the interconnects between arrays of processors, not as actual switches inside the processors themselves. The space constraints in a processor die would make that next to impossible, and at this point there wouldn't be any benefit (a couple inches is a whole lot different than a couple microns).

Though, I recall reading an article about 8 years ago putting forward the idea of of using 'photon channels' inside processors, but the speed of the switch would still be limited by the speed of light, so processor complexity would still be the determining factor in performance.
 
bustaplz said:
I'm no scientist but I thought that wires had resistance?
Click to expand...
Well, you start getting into two different discussions when talking about the "speed" of electricity. And it's more of engineering than science, IMO. But maybe a 50/50 mix of each even.
If you're simply talking about drift velocity, then the speed is directly proportional to the amount of DC voltage. AC voltage really doesn't have a speed, as the electrons don't "move"; they kind of just wiggle (hence your delta-x will be zero and thus velocity will also be zero).
However, if you're talking an electromagnetic wave (which we should be, since that's what actually carries data), then you're talking between 42% and 72% of the speed of light.
Couple of Wiki's:
http://en.wikipedia.org/wiki/Drift_velocity
http://en.wikipedia.org/wiki/Velocity_of_propagation

The thing that I don't seem to follow, and this may be because I honestly have very little clue into how my CPU and all it's interconnects ACTUALLY work is, is that we're talking such short distances, does it honestly matter? 1000X improvement? Color me skeptical!
 
You must log in or register to reply here.
Back
Top