Optical tech.......

[Hellmund]

I'm been hearing a lot lately that's we're nearing the limit on solid state tech, I'm curious on how they would go about using optical technology in processors, anyone any theory's, I was also curious what causes there to be a limit in solid state technology?

[nilknarf]

Solid-State technology (transistors) has a theoretical size limit. Size is a big part of speed. The theoretical size limit for a transistor is around 40 atoms. Unless someone comes up with some very advanced way to make solid state work at smaller sizes, we will hit the size limit in a few years. After that, it's only a matter of time before engineers hit the practical limit on die size. As it stands right now, the Itanium die is so large, it has multiple clocks to keep everything in sync. Keeping the core cool has a big impact on speed too. Superconducting processors operate at near absolute zero. This has the effect of increasing the speed at which electrons can flow through the core. A processor core actually runs slower as it heats up. As engineers design cores with greater energy requirements, they heat up. A P4@1.7G dissipates some 60W of energy in the core, and runs around 100F with cooling. It therefore requires a bigger heatsink and fan. All these factors combine to create a theoretical limit on how fast you can go. I'm not sure what the current theory is about where the top speed is, but I know we're getting there, real fast.

I'll come back later and explain a little more about optical and organic technology.

Tim

Edit - darn KB!

[This message has been edited by nilknarf (edited 05-10-2001).]

[Hellmund]

Looking forward to the other explanations, that's exactly the kind of answer I was looking for, thx a lot.

[DanU]

I recall reading an article about IBM developing a technique to fabricate ~0.01um transistors. This corresponds to about 100 atoms wide. If this becomes commercially viable, we should be able to maintain Moore's Law for at least a decade more, meaning 100GHz processors in 2010.

Here's one article from ZDNET on the subject: http://www5.zdnet.com/zdnn/stories/news/0%2C4586%2C2614200%2C00.html

I haven't heard of any all-optical computers being built, but I have heard of all-optical amplifiers, like the erbium doped fiber amplifier. Computer logic, when looked at from an analog viewpoint, is really just a bunch of inverting amplifiers. So it might be possible to build a computer out of this stuff. This approach to building an optical computer has a long ways to go however, since these devices are extremely large compared to a modern sub-micron transistor.

Personally I'd like to see a revival of bipolar transistor logic. I have a bit of a soft spot for the old BJT. Sure it's considered old school... but it still kicks ***** in the RF world. You may also remember a company called Exponential that fabbed a bipolar processor that was hitting 500-600MHz when CMOS was only at 200MHz.

DIE, CMOS, DIE! CMOS MUST DIE!

[Hellmund]

100ghz........that's dam fast, knowing my luck by the time I get something that powerfull I'll still have this dam M64!

[nilknarf]

It would have to be IBM who comes up with that. They're still a long way from that theoetical limit of 40 atoms.

I'll have to double check, but I think that .01um size would allow transistor densities about ten times greater than current Intel and AMD designs.

Depending on advancements in the manufacturing process, we may hit that mark sooner than that.

I guess come 2010, we'll just have to create a loophole in the law which allows us to do whatever we want!

BTW, when I have more time to spare, I'll talk about the other technologies. And, they are not only possible, but have existed for several years!

Tim

[club_med]

Great info nilknarf, im awaiting your next contribution.

http://www.sysopt.com/forum/smile.gif
cm.

[Hellmund]

How many ~um is 40 atoms?

[Pauly]

I read on the Register a few weeks back that single atom transistors would be possible - they have something to do with 'nanotubes'. I'll see if I can find it again...

paul

[nilknarf]

You have to consider that a transistor is essentially two diodes back to back. This contributes to the size limitation, as you must have enough atoms to control the flow of electrons.

Photons (light), on the other hand, are a completely different story. I'm not sure what the limit is on optical systems, but they will be extremly fast, and probably very expensive.

Dooh! Have to go again.
Tim

[nilknarf]

oops!! double post.

[This message has been edited by nilknarf (edited 05-23-2001).]

[nilknarf]

I just posted some info in this thread:
http://www.sysopt.com/forum/Forum16/HTML/000391.html

Tim

[club_med]

All they have to do is start using different materials for computer components, that should solve the problem for a while.

cm.

[Hellmund]

Sure it's easy to say "just use another material" but just what are they supposed to use? I mean could we see a Silvermine CPU instead of a Cumine?

[nilknarf]

There are only some many materials one can use to make a microprocessor. Intel and AMD just recently (in last year or two) switched for using an 18 micron aluminium interconnects to the 10 to 13 micron copper interconnects. I'm not sure what other materials they may be able to use.

I don't remember if silver and gold are larger or smaller than copper. They are better at transmitting electrical signals, which is why they're used for making high performance connections.

[Axel]

Solid state is limited to the speed of the movement of electrons through a circuit. An electron is a charged particle that has size and other physical factors. We are limited to how close we can place components together before we can no longer insulate the circuit and seperate and control the flow of electrons in the circuit.

Then there's photonics - I.E. light circuits. While light has line of sight issues, it can be reflected and refracted by a number of different means and doesn't have the size limitations that electrons do in a circuit.

Over a decade ago - a photonic processor was created and calculated extremely basic math at an impossibly fast rate.... but it was the size of a large coffee table.... and the trend was that everything was getting smaller and portable..... so photonics didn't go very far because there was still a lot of progress that could be made making electronics smaller and lighter.

So - now we are close to the stage where solid state electron technology has reached it's miniaturization limitations. We'll see more research in photonic research and miniturizing it into stuff that is light enough and mobile enough to be a consumer product.

IBM is working on another concept - quantum technology - I.E., getting away from the electron and going to even smaller nuclear particles to do the computing. As they have to move faster than electrons, we aren't totally certain what the limits of speed are on that or even how to fully utilize it.

I'd say that we will see solid state circuits go away and be replaced by fiber channels even within a PC. You'll see quantum processors and almost everything will have it's own processor - video, audio, com ports, - and in that way free up the bank of main CPU's for other calculations. The days of one CPU on a board are already going away. Now if only code writers would follow the trend and create more multi-processor code out there.