After Die Shrink, whats next?

[joset]

What about ATI?

AM I A BIT MaD

Should it be a question?


Cheers!

 

[joset]

The last M had coNroe envy.

Ah! You're forgiven, then.
On topic, seems a [temporary?] good answer to the thread's title.


Cheers!

 

[computertech82]

Actually the laser is for the prism and will replace hardrives. It exsist today, but too large for a computer tower..so far. One prism can hold the entire library of congress and still have room. The "next generation" cpu will most likey be the bio-chips and bio-memory, the ones the air force has been using for years now for top secret cameras (spy planes).

 

[ches111]

Throwing my "small" hat into the ring ...

What about organics..

Link!

I also think we will get a little smarter/more creative at the use of current technologies.

I think we have only touched the tip of the iceberg in TDM or Time Division Multiplexing. Maybe some hybrid of Time and Wavelength WDM division. Maybe a multiple carrier path with extreme TDM/WDM. I realize this is mostly involving communications right now but it could be applied with adaptive routing algorithms. OC-192 was a dream many years ago but with refinements in fiber purity levels and error retransmission algorithms it is not only possible now, but is fully in use. This is long haul situations which have to worry much more about "loss" than a possible nano-haul situation would have to.

Given current/new technologies "I think" the possibility to do more per cycle will be the next biggest increase we see. Given how good the pre-fetch/predictive branching is getting we are just broadening the use/availability of existing tech.

Some times it may "PAY" more to think/refine inside the box. But greatness/history stands outside of the box.

 

[ches111]

Another thought....

I think with the completion of the ISS International Space Station that we may see some things come to us through the extensive chemical/material testing/experimentation they will be doing. (would love to be a fly on the wall there)

A good GEEK read Properties of materials in space!!

We might get a whole NEW box to think in

 

[kwalker]

I’m amazed at the discussion thus far.
If not for the science fiction theory of electronics we may still hover within caves.
Quantum computing... Lasers… prisms...
Four state???
Why not.
There exists the tri state logic circuit!
Only four would need a controller circuit to function independent bringing us back to two identical devices or maybe one multifunction gateway
With the technology we have available at our fingertips ever expanding theories are fueling the need for a more efficient means.
Perhaps this will all run full circle and bring us back to the abacus.
Personally I believe the future of computing is hidden deep within one of these strange unknown devices.

 

[MU_Engineer]

He was talking about using the transistor where the base voltage was between that required to fully open or shut a transistor- using the transistor as an amplifier over its linear range. Hence the thing about Darlington pair transistors with their ~1000x amplification capability. But that won't give out a binary output and won't work in today's computers.

 

[clue69less]

Another thought....

I think with the completion of the ISS International Space Station that we may see some things come to us through the extensive chemical/material testing/experimentation they will be doing. (would love to be a fly on the wall there)

A good GEEK read Properties of materials in space!!

We might get a whole NEW box to think in

Absativelty! Posalutely! The materials development that has come out of the ISS has already been a bonus. Imagine how different that lab environment will be... Should make for some interesting coatings and adhesives research.

 

[joset]

Given current/new technologies "I think" the possibility to do more per cycle will be the next biggest increase we see. Given how good the pre-fetch/predictive branching is getting we are just broadening the use/availability of existing tech.

Some times it may "PAY" more to think/refine inside the box. But greatness/history stands outside of the box.

I agree and I also believe that's the current view of those involved in Computing technologies. Incremental changes with some unorthodox injects, here & there. Software still has a long way to go and its frontiers are becoming progressively blurred with hardware's.

I would exclude 'zero-gravity' materials' manufacturing from the near-term time span, though, for obvious reasons.


Cheers!

 

[joset]

He was talking about using the transistor where the base voltage was between that required to fully open or shut a transistor- using the transistor as an amplifier over its linear range. Hence the thing about Darlington pair transistors with their ~1000x amplification capability. But that won't give out a binary output and won't work in today's computers.

Yes, a bit like CCDs, these Darlington pairs transistors do increase gain (current density) with relatively low input voltage. Apparently, their advantages compared with more conventional CMOS transistors, end up here:

They provide high current gain (hFE) and require less space than configurations that use two discrete transistors. In Darlington pairs, transistor collectors are tied together and the emitter of the first transistor is directly coupled to the base of the second transistor. The total gain, which is often 1000 or more, is the product of the gain of the individual transistors. Compared to single transistor configurations, Darlington transistor pairs have more phase shift at high frequencies and can become unstable with negative feedback more easily.

(http://transistors.globalspec.com/L...s/Discrete/Transistors/Darlington_Transistors)

I'm sorry to insist but, I still fail to realize where a third intermediate state exists & what purpose it serves, since even Darlington pairs can only be found in two of them: on... & off.


Cheers!

 

[antoant]

Why do you think in states? Why should we consider the transistor either on or off? We could use the transistor in its full operating range and thus use it as an analog device(amplifier?). We could then say that we are going to have x states where each state is a range of voltages.
Example: Each state has a range of 2V thus the first state is from 0-2V then the next state is from 2-4V, the third is from 4-6V etc.
We could then construct an analog computing device. Of course I don't know if we are going to have increased computational power over a device that uses the two logical states but that is not the point. The point is we can make it.

Also since the drain and source of a MOS transistor are exactly the same and thus can be interchanged* we could have a three state device. One state is off, and the other two states are on but with the current flowing in different directions.

*(Take this with a grain of salt, I am not sure if we can do this without any performance hit. Could someone that is more into the physics of the device help us out?)

 

[theaxemaster]

I think, at least for the short term, that you don't even need QM or alternate states. I think tri-gate transistors and FinFET transistors will be the next thing. I know little to nothing of how they work though. I would think you could have 4+ states with them.

 

[joset]

Why do you think in states? Why should we consider the transistor either on or off? We could use the transistor in its full operating range and thus use it as an analog device(amplifier?). We could then say that we are going to have x states where each state is a range of voltages.
Example: Each state has a range of 2V thus the first state is from 0-2V then the next state is from 2-4V, the third is from 4-6V etc.
We could then construct an analog computing device. Of course I don't know if we are going to have increased computational power over a device that uses the two logical states but that is not the point. The point is we can make it.

Yes, I understood it as such (in the case you've presented, 2V is the modulus); but, Im addressing digital devices, not analog.

Also since the drain and source of a MOS transistor are exactly the same and thus can be interchanged* we could have a three state device. One state is off, and the other two states are on but with the current flowing in different directions.

*(Take this with a grain of salt, I am not sure if we can do this without any performance hit. Could someone that is more into the physics of the device help us out?)

Humm... as far as I know yes, source & drain are about the same (electrodes); however, you could not have current flowing both ways simultaneously (and, it also would depend upon nMOS & pMOS...).

Let me put it in more simple terms:

1. Consider a transistor as a switch; you turn on the lights (switch on) or you turn off the lights (switch off);

2. Say the voltage amplitude ("range") between on & off is 1: V ->0, switch is off; V ->1, switch is on;

3. Consider a third ""state"", V=0,5, where the light is dimmed... but still on.

This is my point: Even with a third ""state"", you're still left with lights on or lights off (QC allows a third state: on & off [0, 1] simultaneously; but, that's out of this issue). These, are true states since they imply radical changes in the device's [initial & final] behaviour (transistor). Hence, a third (pseudo) state as you put it, would be identical to the other on state, with two different 'processing' speeds.


Cheers!

 

[Straker]

>2-state devices and logic have been studied in the lab and theoretically for a long, long time.

I remember someone back in the '60s argued 3-state logic would be more efficient than binary logic. Turned out he was wrong.

The reason modern computers use binary-logic circuitry is simply because it's the most economical way to build digital circuits.

 

[joset]

I think, at least for the short term, that you don't even need QM or alternate states. I think tri-gate transistors and FinFET transistors will be the next thing. I know little to nothing of how they work though. I would think you could have 4+ states with them.

Those will be 3D structures (http://www.tgdaily.com/picturegalleries/gallery-20060612-3.html, for instance), a promising approach to improved current flow & electron mobility and leakage decrease; transistor states though, will keep being two: on & off (this starts to be funny; not because n-states aren't possible but, rather, because no-one until now - myself included - wasn't able to define a third transistor state, let alone n...).

JumpingJack is much more knowledgeable than me, on these matters; perhaps it would be interesting to hear his opinion.


Cheers!

 

[antoant]

Ok so we do agree that we can make analog computers.

For the three strate device. I am not saying that a MOS can have current flowing both ways at the same time. What I am saying is that we could have three voltage levels. 0V -3V +3V. Thus if we a capacitor is charged at -3V it is holding state2 if it is charged to 3V is holding state3 and if it is not charged it is holding state1. All the charging is with respect to ground. In this case we don't need transistors to be in a third state. We just need them to be on for only one of the three voltage levels. It does not sound logical to use two state devices to build a three state circuit but it is possible to do so and that is all we are discussing here.

@ Straker: It is more economical to build binary circuits and I am sure it was easier to build them initially because of the simplicity of the logical design. However, we are discussing about the technologies that might be used when we hit the physical wall of Si on which these binary systems are developed. Quantum Computers are not going to be neither simple or cheap. But a lot of people believe that they would be the future of computing. Others say it might be organic computers. Once again those are not going to be simple or cheap. When we find the limit of Si and we cannot push any further then we would have to either stay at that technology level or move to more complex technology. And history has shown that we always go for the second choice.

And to close this, here is an example to counter your example. Back in the day when people were talking about electric cars some dudes said something along these line "Electric cars? Are you insane? I will put a V8 in your car that will make it so fast you will be able to run circles around an electric car going full speed. And it is so much cheaper to build and less complex. Why the hell do you want an electric car? You are stupid to think that electric cars are the way to go." Well guess what. We have reached the limits of the ICE, maybe not because we can't push it any further but because we are kind of running out of oil. So we had to turn to electricity to provide power for our cars. And guess what we have so much more knowledge now about motors and batteries that an electric car can beat a ferrari on a drag race and it can drive just as far on a single charge. Just look up teslamotors on goole.

 

[joset]

I remember someone back in the '60s argued 3-state logic would be more efficient than binary logic. Turned out he was wrong.

Please, care to read the previous posts.


Cheers!

 

[joset]

Ok so we do agree that we can make analog computers.

They started analog!

And, mind this: Wherever "we" go on future computing trends, Quantum Mechanics will be there, as it is now; that's as certain as the Uncertainty Principle!


Cheers!

 

[Straker]

>... but, Im addressing digital devices, not analog. <

You want digital? Okay, consider a 2-bit memory. It can be regarded as a 4-state device. At any given point in time, it will be in one of four possible states.

Physical? Consider a rotary switch with N settings/contacts.

Not trying to argue,... just trying to get at what you are perplexed about.