The end: 5 years

[Chalnoth]

Just got out of my solid state class. It turns out that my professor was involved in some research on single-electron transistors.

Transistors will involve single electrons when their size gets down to 30nm. If transistors per unit area does indeed double every 1.5 years between now and then, we will have hit this barrier within 5 years.

Once you get down to using one single electron per transistor, it should be obvious that you just can't get any smaller without moving to fundamentally different technology.

So, I predict that silicon-based transistors will have hit their end within five years.

What can we do after this? Obviously there are other options.

One obvious one is to move away from silicon. While increased density is still not going to be possible, one can go to, say, Gallium Arsenide (GaAs) as a semiconductor to significantly increase electron mobility. The move to GaAs should increase transistor switching time by a factor of about 5, which should, conversely, increase available clock frequency by a factor of about 5. Many people have been pushing for GaAs in the semiconductor industry for some time, but the companies have always been very reluctant to do so.

Of course, there are other limitations to clock frequency besides just transistor switching speed (which I've stated before). The solution here may be to move to many tiny chips all connected over a lower-frequency bus, though there are very large performance issues to be overcome here.

[DaveW]

Originally posted by Chalnoth
Just got out of my solid state class. It turns out that my professor was involved in some research on single-electron transistors.

Transistors will involve single electrons when their size gets down to 30nm. If transistors per unit area does indeed double every 1.5 years between now and then, we will have hit this barrier within 5 years.


Moores law is slowing down, so we won't continue at a linear rate towards that road block. Transistor count will double in 2 years, then double again in 4, then double again in 8. We have a while yet, but the rate of miniturization will slow. We do have IBMs vertical transistor technology too which can double the density.

Once we reach the point of density, we can just go larger. CPUs and GPUs aren't very big, most of what we see as the "chip" is really its resin housing. I think the size of a P4 wafer today is actually smaller than a 286 wafer years ago. Ofcourse as we go bigger, temperatures go up, but as the fabrication process gets more refined we can get away with running circuits in lower voltages.

I think we have atleast 10 years to go yet.

[Chalnoth]

That slowdown is what I was talking about as "the end."

And yes, it will likely slow down very significantly long before that five year time frame.

But the end result is the same: We are nearly at the limit of silicon-based IC's. It is necessary for there to be a move either to other semiconductors (for higher speed/lower power, but similar densities), or to other technology altogether.

[DaveW]

I like the quantum computer, that can give you the answer to your calculation before you even ask it :)

[Edge]

Oh god, if they can make quantum computers functional to the general public, then it would replace everything else in a second. With speeds the equivalent of 50 gigahertz for the low end, your "advanced" 3.2 gig processor just won't cut it ;)

However, I have heard about something that is supposed to debut in 2007 called the "cell" chip, and I don't know to much about it but I remember hearing that Moore's law does NOT apply to it, so they say that it should be able to reach speeds far above what we currently have. Personally, I think within 10 years computers are going to change radically, almost unrecognizable to us now. We'll all be like our parents who can't get used to computers but know everything about typewrites :D

But we'll have to wait and see, anything can happen in this crazy market, it just depends on how well companies implement thier technology.

[Kruno]

Chalnoth, you know what will be the end?

When machines are able to interface directly with our thoughts. :)

[DaveW]

Or when they start world war III and attempt to irradicate mankind :)

[vampireuk]

and one man is sent back in time to save a woman from been killed by a robot sent by skynet.

Hey this would make a good movie!

[poursoul]

Originally posted by vampireuk
and one man is sent back in time

not a man vampire, but a flibblator.

[Chalnoth]

Originally posted by Edge
Oh god, if they can make quantum computers functional to the general public, then it would replace everything else in a second. With speeds the equivalent of 50 gigahertz for the low end, your "advanced" 3.2 gig processor just won't cut it ;)
I've actually been looking into quantum computers a fair amount. The truth is that they're just not that good for all processing. They're only really good for a few specific algorithms, which can be computed many times faster on a quantum computer than can happen on a silicon-based computer.

The idea really isn't so much about computing faster, but instead about computing differently. This makes available an entire new class of algorithms for computing. It will take a lot of work to figure out how to harness this alternate computing method.

I expect that once these quantum computers become widespread, they will always be used in conjunction with a more traditional processor.

In the meantime, one has to wonder whether or not somebody can come up with a one-molecule transistor or logic gate that can be crystallized and programmed within that crystal to "wire itself" for processing. That's what I see as the holy grail for general-purpose computing.

[Kruno]

Chalnoth, I don't hate electrical goods or anything but when will we see a move from electrical to organic computing?

[Tarin]

Ya, i've got a moldy tomato that i'm not planning to eat...do ya figure if I clean the mold off we can set it up to run DOOM III?

[-=DVS=-]

Organic CPU's as in Startrek ala Voyager had Organic CPU's neural staff , not tomatto lol :lol:
You know human brain does more calculations then all computers combined if we could clone human heads and use them for our games would we sweet :lol:

DVS: hey Gals did you see my new Headputer
Gals: no where is it ?
DVS: its over there in a blue JAR with water :lol:

[Chalnoth]

A few things about biologic processing:

1. It's unreliable. This is a hallmark of life. For harnessing processing capabilities, organic processing would require lots of error correction, wasting much of the processing power (quantum computing has a similar problem).

2. It works at a much higher level than quantum computing. This makes it seem, to me, that organic computing could never attain as high a computing power/volume ratio as quantum computing (speaking purely from logic...I don't know about any existing research on organic computing).

3. Because organic computing works at a much higher level, it stands to reason that it would require a much better understanding of atomic and molecular interactions than we now have. This is why I feel quantum computing will be possible long before organic.

4. Calling it organic really isn't that great of nomenclature. Biocomputing would be better (organic computing, from a technical perspective, would seem to mean carbon-based computing...carbon-based semiconductors certainly are possible, if unfeasible to build).

[Matthyahuw]

bah, who cares what's going on in 5yrs in the computer field, Jesus will be back by then anyway...

[Chalnoth]

Well, I for one, do. This is primarily because I don't think any other computing technologies will be ready to take the forefront within five years. That means the computing industry will slow down, significantly (before it speeds up again).

For part of this, I blame various semiconductor industry giants for vehemently sticking to silicon as the only primary semiconductor. There are better semiconductors.

[Belarnion]

I found this:

http://www.overclockers.com/tips00375/

It's about carbon-based memory.

[tweaked]

Nano-tubes are the future!

[Chalnoth]

Originally posted by Belarnion
I found this:

http://www.overclockers.com/tips00375/

It's about carbon-based memory.
Yes, it would be quite interesting, if they can be built at a reasonable cost. At somewhere around 10 atoms wide, they cannot be easy to build, not by a long shot. As far as I know, nearly all of these technologies are still very much in the research and development phase.