100% Optical computer ?

[runzwithsizorz]

How long before there's a prototype of an optical computer?
Uses protons and light instead of electrons and transistors.

PIC chip has all optical inputs and outputs Exists
Optical modulator Silicon and/ or graphene? Exists
Carbon nanotubes Exists

What are the biggest obstacles in building one?( Not counting cost)

The Wife

 

[videogames101]

optical signaling is easy enough, but what do you use to hold and retrieve state?

(honest question)

 

[inachu]

optical signaling is easy enough, but what do you use to hold and retrieve state?

(honest question)

A light based transistor. Or not everything needs to be based on light.

Crystals already show they can hold certain information.
These crystals will need to be man made so that when grown they will be preset to store data better than in their natural state.

 

[bryanl]

PIC chip has all optical inputs and outputs Exists

What is the model number? I think it would use optical I/O for noisy environments or galvanic isolation, not speed.

 

[John Connor]

I have always wondered this. Would have some interesting applications.

 

[dave_the_nerd]

http://www.technologyreview.com/tagged/optical-computing/

Interesting reading. Sounds like it's been ten years off for twenty years or so. Oh well.

 

[Cogman]

Probably never.

The first question to ask is "What are the advantages of an all light circuitry". Answer? Very little. Transistors are very small and work very fast. Light transistors haven't proven to be the same. In fact, they are currently large, slow, and mostly theoretical. So at the basic logic level, electrical transistors in whatever form will more than likely dominate. The only way light will take them out is if through some form of magic the light transistor ends up smaller and faster than regular transistors.

That said, there are some clear advantages to using optics in communication which is why the likes of Intel are looking on incorporating them into their CPUs. The biggest advantage is that light doesn't interfer. It doesn't induce currents on surrounding wires, a light wire could pass right next to any part of the cpu and not interrupt it. That may make it possible for less layers for the CPU which translates to higher yields. A vast majority of CPU layers are currently dedicated to wiring. It also may free some design constraints.

What it doesn't do (necessarily) is improve speed or power consumption.

 

[William Gaatjes]

Probably never.

The first question to ask is "What are the advantages of an all light circuitry". Answer? Very little. Transistors are very small and work very fast. Light transistors haven't proven to be the same. In fact, they are currently large, slow, and mostly theoretical. So at the basic logic level, electrical transistors in whatever form will more than likely dominate. The only way light will take them out is if through some form of magic the light transistor ends up smaller and faster than regular transistors.

That said, there are some clear advantages to using optics in communication which is why the likes of Intel are looking on incorporating them into their CPUs. The biggest advantage is that light doesn't interfer. It doesn't induce currents on surrounding wires, a light wire could pass right next to any part of the cpu and not interrupt it. That may make it possible for less layers for the CPU which translates to higher yields. A vast majority of CPU layers are currently dedicated to wiring. It also may free some design constraints.

What it doesn't do (necessarily) is improve speed or power consumption.

That light wave propagating through the die. That would make an interesting hybrid chip. Mayby light as clocksignal. Light also saves in energy for interconnects as IO between chips.

 

[Cogman]

That light wave propagating through the die. That would make an interesting hybrid chip. Mayby light as clocksignal. Light also saves in energy for interconnects as IO between chips.

I'm not entirely convinced it saves energy between chips, but even if it does, I'm not convince that chip to chip IO has ever been a major power consumer. It certainly allows for closer packing and smaller packaging as the light tunnels can be right on top of themselves.

 

[WhoBeDaPlaya]

That light wave propagating through the die. That would make an interesting hybrid chip. Mayby light as clocksignal. Light also saves in energy for interconnects as IO between chips.

Silicon being an indirect bandgap material ups the cost a bit, not to mention that fabrication has to be even more tightly controlled (angle of the waveguide sidewalls, RMS roughness, etc.)

 

[MongGrel]

I'd think something along those lines would be in the works.

optical signaling is easy enough, but what do you use to hold and retrieve state?

(honest question)

I believe IBM was working on a crystal cube type of thing years back for storage, I do not have a book mark of it offhand.

 

[inachu]

I'd think something along those lines would be in the works.



I believe IBM was working on a crystal cube type of thing years back for storage, I do not have a book mark of it offhand.

In helping optics out we have this:
http://www.theregister.co.uk/2014/10/01/xmen_boffins_demo_nanomagnets_to_replace_transistors/