The future of CPU's

[wolfman140]

So as I read the IDF article, this one just published and previous, I see that Intel is begining to experiement with laser/light transmission for processors rather than traditional means. Apparently they said if successful it will allow much faster transfer rates and speeds. I also wonder, if this will mean that it won't need nearly as much cooling? I suppose you still need to power the lasers, but the overall heat generated would be less than today's procs yes?
Anyway...I think this is a very exciting time..With the quads coming available already in November! And did you see that bloody 16-core computer? Wow. What do you use, and how, do you connect 4 motherboards together? I take it they are server motherboards with a specific port to do this yes?
What's eveyone's thoughts on this? Do you think these laser procs will replace traditional ones? 10 years? 5 years? Do you think the proc industry is moving WAAAYY to fast for the software writers to catch up?

 

[rush_123]

i think its about time they got something working. i am supprised they still don't have the whole chip working on light.

 

[SciFiMan]

Maybe 15-20 years out for a laser based CPU, probably longer for consumers. We'll see them first in very specialized chips such as for communications. Then they'll trickle down into the high end servers. Maybe consumer CPU's, a long time from now. The PC would become to unbalanced. First we'll need standards for a new motherboard design, interconnects, bus, memory, graphics, etc. Not to mention much faster disks, or the demise of mechanical disks and move us onto 500GB flash drives. A super fast CPU won't mean much if the rest of the system is 100x slower.

Software developers don't have to do much worrying about hardware. Almost nobody writes driectly to hardware anymore (very specialized cases). Any program will just run faster like it does now when moving to a faster CPU.

 

[thelostchild]

Software developers don't have to do much worrying about hardware. Almost nobody writes driectly to hardware anymore (very specialized cases). Any program will just run faster like it does now when moving to a faster CPU.

and even then wouldn't it just be like moving to a new archetecture because the processor must still use logic gates etc. like it does today?

 

[FeareX]

in the future, we dont measure the cpu's performance in GHz but in number of cores. :lol:

 

[mr_fnord]

That's the problem with optical computing, logic gates don't work the same with light, since one beam of light doesn't affect another. Optics are great for communication, since there is no voltage loss, interference, etc. but until someone creates an all optical logic gate, and one that runs at Ghz processor speeds, optical systems will sit on the sidelines.

Replacing buses with optical pathways might be valid, but you still have to contend with the conversion to and from optical signals, and I would be currious to hear what a 6 layer PCB made with fiber optics instead of traces would cost.

Now, optical computing along the lines of quantum computing might be possible, taking advantage of the unique physical properties of light just as a transistor takes advantage of the unique characteristics of electricity. That, however, would be a major paradigm shift.

 

[joset]

Like mr_fnord said, near-term optical devices will be used for comms & system interconnects; optical computing ("optical" processors) are still far from a< desktop near you.


Cheers!