US Intel Agencies To Build Superconducting Computer

dcblogs writes "The Director of National Intelligence is soliciting help to develop a superconducting computer. The goal of the government's solicitation is 'to demonstrate a small-scale computer based on superconducting logic and cryogenic memory that is energy efficient, scalable, and able to solve interesting problems.' The NSA, in particular, has had a long interest in superconducting technology, but 'significant technical obstacles prevented exploration of superconducting computing,' the government said in its solicitation. Those innovations include cryogenic memory designs that allow operation of memory and logic in close proximity within the cold environment, as well as much faster switching speeds. U.S. intelligence agencies don't disclose the size of their systems, but the NSA is building a data center in Utah with a 65 MW power supply."

Can superconductors compute?

[timeOday]

I am about to ask a very naiive question so please bear with me. Interconnects aside, is an ideal transistor permitted by theory? That is, 0 resistance when closed and "infinite" resistance when open? (Surely not the latter, since arcing could occur even in a vacuum). And while we're at it, it should not require any current to hold the transistor open or shut once it is switched. And should be infinitely fast :)

There must be a divide by 0 in there somewhere, it just doesn't seem like the universe would permit computation without creating some entropy.

Re:Can superconductors compute?

[Anonymous Coward]

You don't need infinite resistance insulators if there is a path with 0 resistance (as long as its not saturated). Also, there are way better insulators than vacuum.

However there always will be some losses: if you want to represent a bit, it must require some switching energy, or it will thermally get switched. This is where the massive gain from being very cold comes from: you can have way lower energy bit representations.

There is also always some capacitance, and connecting a low bit to a high bit is much like connecting capacitors together: you lose half the energy no matter how low the resistance is (assuming one is charged and the other is not, and they have equal capacitance, you get half the voltage across double the capacitance, and due to the V^2 term, half the energy).

So, you can't do an ideal job with transistors because changing bits the way they do consumes energy. But, it seems likely that if you managed to make super conducting transistors at very low temperatures, you might be able to drastically reduce the losses and / or run much faster. That is what they are aiming for.

You might even be able to use smaller transistors to have more computer per area, since the bits need to store less total charge with the lower thermal noise. If you throw in error correction to deal with the occasional cosmic ray or other random effects, you might be able to push the size of the stored charge way down from current approaches.