Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Virex]

I meant that making them in the first place is easier, or so I recall, ignoring what happens after that, because nothing prevents you from coating the coper wires using MBE or something.

[TheBronzePickle]

Unless said wires can't be coated because of their function or a need to use as little physical space as possible.

[Virex]

Unlikely, as a coating needs only be a few atoms thick, while a nanowire is usually on the scale of 20 to 50 nanometers. If you go below that, the conductivity of metal nanowires drop steeply due to the lower mean free path of electrons and the effect of surface dislocations and edge effects. If you really need conductivity on a smaller scale you'd use carbon nanotubes instead, which don't suffer from these problems. They can go as small as 0.5 nm diameter, which is pretty much the lower limit at which you can still engineer any structures.

[Montague]

I thought fiber optics were supposed to replace these kinds of miniaturized conductors anyways.

[TheBronzePickle]

Fiber optics isn't efficient in a computational circuit.

[Virex]

It's not really my field, but I think it's unlikely that optical computing will replace semiconductors soon. The problem is that right now we can't make nanoscale lasers, you need at least a 300 nm cavity and several millimeters of photonic crystal around that. That alone puts a hard cap on the lowest size components you can get. That means to get the same amount of optic switches on a chip as we currently have, you'll need a room-sized computer. Of course that comparison isn't justified because optic computers are faster, but it's still a factor 100 in size they need to make up for. I could easily imagine it being worth it for routers and maybe supercomputers, but not for laptops or desktops. Unless metamaterials or plasmon-based lasers take off much faster than expected we won't see optical computers in home applications for the coming decades, maybe even longer. And it's still competing with ever-smaller semiconductors, which leech of the same technologies.

[Eagleon]

It's not really my field, but I think it's unlikely that optical computing will replace semiconductors soon. The problem is that right now we can't make nanoscale lasers, you need at least a 300 nm cavity and several millimeters of photonic crystal around that. That alone puts a hard cap on the lowest size components you can get. That means to get the same amount of optic switches on a chip as we currently have, you'll need a room-sized computer. Of course that comparison isn't justified because optic computers are faster, but it's still a factor 100 in size they need to make up for. I could easily imagine it being worth it for routers and maybe supercomputers, but not for laptops or desktops. Unless metamaterials or plasmon-based lasers take off much faster than expected we won't see optical computers in home applications for the coming decades, maybe even longer. And it's still competing with ever-smaller semiconductors, which leech of the same technologies.

Correct me if I'm wrong, but isn't one of the biggest potential strengths of optical computers the fact that you can use optical amplifiers and beam-splitting to get as many streams going as you like? I thought the idea was to only use one laser, a few at most, and put it through non-laser components that modify the signal the way a processor would. So you have your laser pumping out light at the top (the 'light supply' next to the power supply maybe), a set of beam-splitters multiplying your laser's path (and if necessary amplifiers to boost intensity), and gates into your optical processor's operation paths switching on and off in pattern. I know it's not as simple as this but I never thought of the bulk of the components as an obstacle, merely designing such a system and figuring out silicon-based components that can replace conventional logic gates.

[TheBronzePickle]

One issue is that the light's transport paths are going to have to be at least as large as the wavelengths of the light, which is pretty big unless we start getting into ionizing radiation that could potentially endanger lives. Electricity can flow across smaller paths.

[Pnx]

Would it be possible to do ternary computing with light based computers if we ever adopt them?

I know we're unlikely to see either of them coming along any time soon, but it's a nice thought.

Well heck, pretty much all the ideas discussed here are unlikely to see implementation any time in the next 50 years. It's less an issue of technology more an issue of politics and logistics.

[Montague]

Well, if fiber optics are faster and more efficient, there really might not be any need to keep miniaturizing them like with conductors. Even if the components are limited by the practicality of physics to be bulkier they still might be very powerful for any computer the size a person might practically use. They could just be made as compact as possible and built into modules that can be interchanged to scale up the size and computing power of the device.

Or I might have no idea what I'm talking about here.

[Eagleon]

One issue is that the light's transport paths are going to have to be at least as large as the wavelengths of the light, which is pretty big unless we start getting into ionizing radiation that could potentially endanger lives.

You underestimate the lengths to which I and others will go to play Dwarf Fortress 2030

[TheBronzePickle]

Electricity travels at the speed of light, or at least very close to it. The problem is with switches, and those can be and are being made to move faster. Circuit-based computing is still the major focus of processing, and until its development stops for whatever reason it's probably what processors are going to be based on well into the future.

That doesn't mean we can't make computers faster by tying the processors together with things like optics, however, so it will still be an important part of computing.

[Virex]

On a chip it goes at roughly half the speed of light, because there's a medium slowing down the propagation of the electric field. Of course light doesn't go at the speed of light in an optical circuit either.

It's not really my field, but I think it's unlikely that optical computing will replace semiconductors soon. The problem is that right now we can't make nanoscale lasers, you need at least a 300 nm cavity and several millimeters of photonic crystal around that. That alone puts a hard cap on the lowest size components you can get. That means to get the same amount of optic switches on a chip as we currently have, you'll need a room-sized computer. Of course that comparison isn't justified because optic computers are faster, but it's still a factor 100 in size they need to make up for. I could easily imagine it being worth it for routers and maybe supercomputers, but not for laptops or desktops. Unless metamaterials or plasmon-based lasers take off much faster than expected we won't see optical computers in home applications for the coming decades, maybe even longer. And it's still competing with ever-smaller semiconductors, which leech of the same technologies.

Correct me if I'm wrong, but isn't one of the biggest potential strengths of optical computers the fact that you can use optical amplifiers and beam-splitting to get as many streams going as you like? I thought the idea was to only use one laser, a few at most, and put it through non-laser components that modify the signal the way a processor would. So you have your laser pumping out light at the top (the 'light supply' next to the power supply maybe), a set of beam-splitters multiplying your laser's path (and if necessary amplifiers to boost intensity), and gates into your optical processor's operation paths switching on and off in pattern. I know it's not as simple as this but I never thought of the bulk of the components as an obstacle, merely designing such a system and figuring out silicon-based components that can replace conventional logic gates.

As said, it's not my field so I may be mistaking some things. However, all components working with light have to be roughly the size of the light's wavelength or more. That includes the switches, memory units and as said the waveguides. Therefor, for computation, optic chips are just too bulky. But as said, it's certainly an option for routers.

[palsch]

Unless said wires can't be coated because of their function or a need to use as little physical space as possible.

My experience of this is using a capping layer of gold over the top of any exposed circuits. Of course, this was using iron compound wires (FeNi permalloy or Fe₃O₄ magnetite) built on top of wafers for lab and imaging purposes. Working with commercial designs it's easier to close the whole system. Especially when you are talking about such a small system as in a computers processor or similar. Even if you did go with capping layers, the thickness would be negligible (a few atoms).

As for optical computing or any other form, I doubt any will catch up in the next decade. The latest chip architecture uses a 22nm transistor, with 14 and 10nm versions on the near horizon. For computation, size is speed. Until you hit some threshold (I'd guess for electric transistors it would be around 5nm or so, but potentially anywhere under 10nm) when things stop working right, smaller is better.

There just aren't any realistic competitors for this yet.

I don't see any demands from computing that would make space resources a sensible goal. Then again, I don't see any earth based commercial or industrial interests that would. Lunar resources would make sense for a moon based industry, which makes marginally more sense than an earth based one. But then, once you are outside the local gravity wells you may as well stay there.

[Virex]

Except that if you're outside of a gravity well, you lose about 80% of our current purification technology and what remains either costs a lot of energy to operate or just plainly doesn't work for most of the things you'd find in outer space.