Would it be possible to make the wiring inside a chip out of niobium or some other good metal supercondutor then mate it with traditional transistor devices? Wouldn't this allow for vastly increased clock speeds, and nearly 0 heat output? If not, what keeps such a thing from happening?
Super conducting chips
- Thread starter FishTankX
- Start date
HI - a super conductor only acts as a super conductor at very low temperatures, not far off 0 Kelvin (-273 degrees centigrade) - hence the chip would have to be very very cold for it to work 
There is a of research into superconducting electronics. As was pointed out above the problem is that superconductirs only work below a certain temperature. Niobium becomes superconducting at 9.2K. There are a few places in to world that can make niobium chips with around 50 000 elements.
The basic building-block in superconducting electronics is not a transistor, instead something calles a Josephson junction is used. This technology is known as RSFQ (rapid single flux quantum). RSFQ-chips can have clock-frequencies of many tens of GHz (many prototype circiuits operate at 80-90 GHz).
High-Temperature superconductors become superconducting at higher temperatures, YBCO becomes superconducting at 92 K but the manufacturing process is much more complicated and so far we have only been able to make circuits with around 100 elements. This is still be enough for certain applications in telecommucations. 92K is still cold but there are cheap cryocolers (cryocolers are well known to overclockers since they can be found in many cooled cases) that can cool the chips to around 30 K.
Superconducing electronics is used commercialy in the telecommunications industri, for example in some base-stations for mobile phones.
/Tobias
The basic building-block in superconducting electronics is not a transistor, instead something calles a Josephson junction is used. This technology is known as RSFQ (rapid single flux quantum). RSFQ-chips can have clock-frequencies of many tens of GHz (many prototype circiuits operate at 80-90 GHz).
High-Temperature superconductors become superconducting at higher temperatures, YBCO becomes superconducting at 92 K but the manufacturing process is much more complicated and so far we have only been able to make circuits with around 100 elements. This is still be enough for certain applications in telecommucations. 92K is still cold but there are cheap cryocolers (cryocolers are well known to overclockers since they can be found in many cooled cases) that can cool the chips to around 30 K.
Superconducing electronics is used commercialy in the telecommunications industri, for example in some base-stations for mobile phones.
/Tobias
My question is this. Would it be possible to make a traditional transistor based circuit with super conducting wiring, and then cool it down to superconducting levels? I know about super conductors and how you have to cool niobium with liquid helium and all of that stuff. I know super conductors pretty well.
Let me ask you another question then. If you made a traditional transistor circuit out of super conductors, would it consume any power? If so, wouldn't it be significantly less than a normal copper based circuit?
Let me ask you another question then. If you made a traditional transistor circuit out of super conductors, would it consume any power? If so, wouldn't it be significantly less than a normal copper based circuit?
It is possible to make "conventional" electronic circuits using Nb. In fact it was done some 20 years ago. The first superconducting microprocessor was built in 1984(I think , might have been 1983) using aluminium (superconducting below 1.2K).
Even a superconducting processor will consume power simply because you have to switch something in order to do something and the switching produces heat. It will consume much less power than a copper circuit (interestingly enough it ca also be shown that it is impossible to "compute" without producing heat, moving bits around will always produce heat no mather what kind of processor you use).
The reason why it is not done it simply that there is no reason. RSFQ is much faster and simpler and you might as well use it if you are going to use superconductors. I should point out that even RSFQ produces some heat.
/Tobias
Even a superconducting processor will consume power simply because you have to switch something in order to do something and the switching produces heat. It will consume much less power than a copper circuit (interestingly enough it ca also be shown that it is impossible to "compute" without producing heat, moving bits around will always produce heat no mather what kind of processor you use).
The reason why it is not done it simply that there is no reason. RSFQ is much faster and simpler and you might as well use it if you are going to use superconductors. I should point out that even RSFQ produces some heat.
/Tobias
P.S. If this is all true, then why not supercool computers down to the superconducting level to get the most power efficency out of them? It'd seem that crycooling them down to that point might actually save power bills.
well you're already paying absurd amounts of money for airconditioning, running fans, and you'd simplify system designs by doing away with cooling constraints, which might allow more design flexibility.
Whatever works.. but it'd seem like you'd save alot of money by doing away with airconditioning. Espically by making the wirest out of high K superconductors, if that would be possible. That, plus the massive electricity savings from the superconducting carriers..
Whatever works.. but it'd seem like you'd save alot of money by doing away with airconditioning. Espically by making the wirest out of high K superconductors, if that would be possible. That, plus the massive electricity savings from the superconducting carriers..
Isn't reversible computation meant to take up no heat?
It was my understanding that moving bits around consumes negligible amounts of heat and only creating or destroying bits make heat. I think some guys have made a completely reversible computer and it puts out like a couple of milliwatts but is also painfully slow.
It was my understanding that moving bits around consumes negligible amounts of heat and only creating or destroying bits make heat. I think some guys have made a completely reversible computer and it puts out like a couple of milliwatts but is also painfully slow.
Reversible computing is very strange. As far as I can remember it uses so-called Toffel-gates which preserves all information about the input to the gata (that is, in the instruction x OR y which gives the output TRUE or FALSE it would also preserve the values of x and y). Reversible computers are not very effiecient. Toffel-gates (I think that is the name) are most known for the fact that they are the basic gates in quantum-computers.
About power-consumtion: I heard a talk today in which it was stated that you could reduce the power consumtption to half in a high-end computer system if you used superconducting technology, I don't know how they calculated that figure. The thing is that even in a superconducting chip you have normal-conducting components, in fact the operations are done by switching the elements (the Josephson junctions) into the normal state for a brief moment which means that it will produce heat.
And finally about superconduting wires made from high temperature superconductors: It is in principle possible but it is very difficult to connect normal-metal and a superconductor such as YBCO (superconducting at 92K) without huge high-frequencie losses (DC is easier and that is why it is used to transfer high currents). It is a technical problem and it might be possible to solve but since most of the power is dissipated in the transitors anyway I am not really sure it would be worth it.
About power-consumtion: I heard a talk today in which it was stated that you could reduce the power consumtption to half in a high-end computer system if you used superconducting technology, I don't know how they calculated that figure. The thing is that even in a superconducting chip you have normal-conducting components, in fact the operations are done by switching the elements (the Josephson junctions) into the normal state for a brief moment which means that it will produce heat.
And finally about superconduting wires made from high temperature superconductors: It is in principle possible but it is very difficult to connect normal-metal and a superconductor such as YBCO (superconducting at 92K) without huge high-frequencie losses (DC is easier and that is why it is used to transfer high currents). It is a technical problem and it might be possible to solve but since most of the power is dissipated in the transitors anyway I am not really sure it would be worth it.
Liquid nitrogen has to come from somewhere, and as somebody mentioned already, HTS material is harder to produce than LTS material, and is more susceptible to stray magnetic fields anyway.
Liquid helium costs much more than liquid nitrogen, so what makes you think you could save on energy using it?
Liquid helium costs much more than liquid nitrogen, so what makes you think you could save on energy using it?
What makes me think we could save money on doing it, is because if you cut the energy of the chip in 1/2, you cut the energy needed for airconditining and crap to 1/4, since generally you get a 2/1 ratio. Whether or not it's worthit, I don't know. I don't know for sure the efficencies of super conducters in circuitry.
I do know that with crystal clear lines (super conductors are always 100% efficnent) you could run them on ultra low voltage (Because not much power is needed to distinguish between 0 and 1, if you're using traditional electronics) and josephson junctions are hard to build so it might be easier just to bring the temperature of traditional conductors down to that level.
But the thing that intrigues me is that generally semiconductor performance goes up as temperature goes down. It might make it possible to squeze as much processing power out of 1 chip as you might get out of 4 in the first place. Super high frequencies, lower power (Less voltage), etc.. would equal one mean super computer.
I do know that with crystal clear lines (super conductors are always 100% efficnent) you could run them on ultra low voltage (Because not much power is needed to distinguish between 0 and 1, if you're using traditional electronics) and josephson junctions are hard to build so it might be easier just to bring the temperature of traditional conductors down to that level.
But the thing that intrigues me is that generally semiconductor performance goes up as temperature goes down. It might make it possible to squeze as much processing power out of 1 chip as you might get out of 4 in the first place. Super high frequencies, lower power (Less voltage), etc.. would equal one mean super computer.
It still sounds like you'll still need a hell of a lot more energy to compress the helium than you save from a piddly 100W of consumption.
Ah, well, how much do you think they pay for the airconditioning and the power that they'll save from having perfect conductance in the chip's innards? Heh. Alot.
I'm just thinking it's not just the energy saving factor, you know. It's more of the performance factor that intrigues me. How low do you think the voltage could go on a P4 if you used super conductors? Who's to say that you couldn't take down the voltage as low as she'll go, because you never have to worry about noise in the circuits?
I'm just thinking it's not just the energy saving factor, you know. It's more of the performance factor that intrigues me. How low do you think the voltage could go on a P4 if you used super conductors? Who's to say that you couldn't take down the voltage as low as she'll go, because you never have to worry about noise in the circuits?
Liquid helium has been used for a long time to cool semiconductors. Especially in detector applications, for example in astronomy and the performance goes up (basically since the noise-level goes down).
One problem is cost, if you cool a normal silicon-chip to helium temperatures it will stop working since the charge carrier "freezes out" (the silicon becomes insulating). At helium-temperatures you have to use gallium-arsenid or gallium-nitride.
The problem is cost: the price is about $13/dm^3 where I live.
Liquid nitrogen is much cheaper, it is less $1/dm^3.
One problem is cost, if you cool a normal silicon-chip to helium temperatures it will stop working since the charge carrier "freezes out" (the silicon becomes insulating). At helium-temperatures you have to use gallium-arsenid or gallium-nitride.
The problem is cost: the price is about $13/dm^3 where I live.
Liquid nitrogen is much cheaper, it is less $1/dm^3.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 24K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter