Moore's law again, I'm nothing if not predictable ;)

[jasong]

I've been thinking about the possible future of the not-truly-a-law Moore's law, and I realized something.

Even if we hit a barrier where we can't shrink things down or improve speed, there's still efficiency and cost. If the capabilities of chip stay exactly the same, but the manufacturing costs go way down for whatever reason, like stagnation(hopefully, temporary) of the industry, if that's what happens( run-on sentence for the win) then it could still be considered as operating within Moore's Law. My interpretation of the law, fwiw, is the whole concept of double the awesomeness for the same price.

So if only the price moves, it still fits with that. Additionally, there's also specialization, like chips and GPUs made for highly specific workloads.

Lastly, we could have huge chips(relative to what we're used to) that cover everything but the kitchen sink, where a lot of the basic architecture is repeated multiple times and what is used is dependent on what the specific workload is.

Lots of possibilities, even if technical advancement comes to a grinding halt.

[jasong]

from Wikipedia:

In 2003, Intel predicted the end would come between 2013 and 2018 with 16 nanometer manufacturing processes and 5 nanometer gates, due to quantum tunnelling, although others suggested chips could just get bigger, or become layered.[58] In 2008 it was noted that for the last 30 years it has been predicted that Moore's law would last at least another decade.

We have a couple possibilities for the extension of Moore's Law after the atomic limit is reached.

The first possibility is gallium arsenide(not sure if that's the right term) which may be able to get chips up to 100GHz. The reason we don't have it yet is the production process produces arsenic gas, so China might be able to get away with making it, but nobody else.

The other possibility is photonic transistors. But I'm not sure how useful they would be on an actual microchip, they might simply be used for communication purposes, faster switching at internet relay stations. So less lag on the Internet.

A third possibility is carbon nanotubes. The appeal of carbon nanotubes is that electrons are a lot less likely to go where you don't want them to go, so you'd get less accidental heating of the microchip, which is always good.

Thoughts?

[ixfd64]

Quantum computers?

[bsquared]

GaAs is not a path to continuation of Moore's law. GaAs is mostly used for constructing HEMT devices, which are fast but relatively high power devices that can't be used for VLSI. It is primarily used in radar, EW, low noise receivers and the like.

[retina]

I liken the end of Moore's Law to the situation with cars today. Once the circuitry cannot be shrunk any further then it will simply become a matter of designing chips to suit the market, rather than the market changing to suit the available chips like we have today.

BTW: QCs are not a solution to "solve" Moore's Law's demise.

[LaurV]

Originally Posted by bsquared

GaAs is not a path to continuation of Moore's law. GaAs is mostly used for constructing HEMT devices, which are fast but relatively high power devices that can't be used for VLSI. It is primarily used in radar, EW, low noise receivers and the like.

We are actually witnessing a retro-movement (or going-back? how it is called?) to germanium, which was used 50 years ago before the silicon became so popular. Already we have the 30GHz to 40GHz transmitters (for wireless, tv satellites, eves some industrial stuff which before was done with mechanical waveguides exclusively) which are done now with germanium. With the existent manufacturing technology of today, we can make integrated circuits with germanium which are compatible as power consumption with the one with silicon, but much faster (also much bigger, but I imagine the things will go in the right direction after someone makes first steps...). Even NASA put it in their spaceships... One reason why the process is slow is the higher costs of the germanium, and the technology not so developed, up to now it was (and it will be for the next 10-20 years or more) much cheaper and more efficient to use silicon.

[bsquared]

Originally Posted by LaurV

We are actually witnessing a retro-movement (or going-back? how it is called?) to germanium, which was used 50 years ago before the silicon became so popular. Already we have the 30GHz to 40GHz transmitters (for wireless, tv satellites, eves some industrial stuff which before was done with mechanical waveguides exclusively) which are done now with germanium. With the existent manufacturing technology of today, we can make integrated circuits with germanium which are compatible as power consumption with the one with silicon, but much faster (also much bigger, but I imagine the things will go in the right direction after someone makes first steps...). Even NASA put it in their spaceships... One reason why the process is slow is the higher costs of the germanium, and the technology not so developed, up to now it was (and it will be for the next 10-20 years or more) much cheaper and more efficient to use silicon.

Well, he asked about gallium arsenide, not germanium, but I agree that germanium (actually, SiGe) is very competitive with and has distinct advantages over GaAs. SiGe BiCMOS processes in particular are really cool, coupling CMOS logic with high speed SiGe HBTs. But the level of integration is still comparatively low, and the high speed HBTs are still primarily used in communication devices. You will not see 2 billion SiGe HBTs on a chip anytime soon.

[davieddy]

Originally Posted by bsquared

gallium arsenide

Shouldn't this belong in the Soapbox?

[jasong]

Originally Posted by davieddy

Shouldn't this belong in the Soapbox?

I've always thought of the Soapbox as a place for stuff that's potentially offensive to people. Unless we start fighting, I think it's where it should be.