80 Times More Energy Prototype Superconducting Chips that Will Soon Reach 10 Ghz

Researchers from Yokohama National University in Japan have developed a prototype microprocessor using superconductor devices that are about 80 times more energy-efficient than CMOS semiconductor devices. Computers currently use 10% of global electricity.

The adiabatic quantum-flux-parametron (AQFP) is a building block for ultra-low-power, high-performance microprocessors, and other computing hardware for the next generation of data centers and communication networks. They made a four bit AQFP that proves practical energy-efficient high-speed computing is possible. The prototype 4-bit AQFP microprocessor called MANA (Monolithic Adiabatic iNtegration Architecture) is the world’s first adiabatic superconductor microprocessor.

A separate chip shows data processing part of the microprocessor can operate up to a clock frequency of 2.5 GHz which is around the level of 2-4 Ghz par for current CMOS computing technologies. They expect this to increase the AQFP chip clockspeeds to 5-10 GHz with some modifications.

The devices have to be cooled 4.2 Kelvin to allow the AQFPs to go into the superconducting state.

IEEE Journal of Solid State Circuits – MANA: A Monolithic Adiabatic iNtegration Architecture Microprocessor Using 1.4-zJ/op Unshunted Superconductor Josephson Junction Devices

Abstract:
We conducted the first successful demonstration of an adiabatic microprocessor based on unshunted Josephson junction (JJ) devices manufactured using a Nb/AlOLₓ/Nb superconductor IC fabrication process. It is a hybrid of RISC and dataflow architectures operating on 4-b data words. We demonstrate register file R/W access, ALU execution, hardware stalling, and program branching performed at 100 kHz under the cryogenic temperature of 4.2 K. We also successfully demonstrated a high-speed breakout chip of the microprocessor execution units up to 2.5 GHz. We use a logic primitive called the adiabatic quantum-flux-parametron (AQFP), which has a switching energy of 1.4 zJ per JJ when driven by a four-phase 5-GHz sinusoidal ac-clock at 4.2 K. These demonstrations show that AQFP logic is capable of both processing and memory operations and that we have a path toward practical adiabatic computing operating at high-clock rates while dissipating very little energy.

SOURCES- Eurekalert, Yokohama National University,IEEE Journal of Solid State Circuit
Written By Brian Wang, Nextbigfuture.com

23 thoughts on “80 Times More Energy Prototype Superconducting Chips that Will Soon Reach 10 Ghz”

  1. Cooling costs have already been considered.

    from the link above:

    VII. CONCLUSION We have designed and demonstrated an adiabatic micro-processor called MANA using unshunted superconductor JJ devices. We implemented this microprocessor using AQFP logic. The basic AQFP cell has a measured switching energy of only 1.4 zJ per JJ at 4.2 K. Even after considering cooling costs, the switching energy is approximately only 1.4aJ per JJ.

  2. Helium boils at 4K so just keep pouring it on, easy, but very expensive.

    Hydrogen boils a 20K, that would be a lot cheaper.

  3. As I learned, soon after making the comment.

    I started reading about the topic from other sources, then looked into superconductive monopolar DC motors.

    Very interesting stuff.

  4. Yes, these designs are for *normal* cooling. But they are *free* in a way, and can then be the sinks for powered refrigeration on down.

    "The DIVINER instrument on LRO has measured the surface temperatures of
    the polar regions and find these dark areas to be extremely cold, never
    reaching surface temperatures greater than about 25° to 30° above
    absolute zero (–272° C). " But still a long way to go. James Webb tested to ~27K, if that means anything. I know my network guy would be happy with a lot less.

  5. 4.2K is probably too low a temperature to really get any significant amount of radiative cooling. 4th power law, after all.

    The temperatures on Pluto, around 50K, probably demonstrate the lower limit of radiative cooling given any degree of heat generation.

  6. They've been applying it to integrated circuits for a long, long time. The problem is that this degree of cooling is expensive, and you start to get serious stresses in your circuitry due to differential thermal contraction.

  7. Two biggest problems I see are (1) this is only a 4bit CPU core running at 100kHz. The 2.5GHz part was a test structure…. and (2) unless they've found a way to eliminate capacitance on the interconnect, then most of the power will still be there. Even if the transistor (and routing) has no resistance, there is still capacitance which must be charged and discharged identically to a classical computer.

  8. IR (edit: should have been more general, "black body radiation") radiation is the *true* radiation. The other stuff called "radiators" are usu heat exchangers. They dissipate by contact, so fans help. Space IR is *seeing* 2-3 K temps to cool *into*. Now, the problem is spreading the heat out so a large part of the dark Space is avail to accept it. Big tubes with particles slowly blowing thru so the particles touch the cold tube have been proposed. A huge topic if there is no water or air to dump heat into.

  9. "The devices have to be cooled 4.2 Kelvin to allow the AQFPs to go into the superconducting state."
    -> Ah, OK
    I am not so sure a refrigeration element that can cool down to 4.2K is that practical

  10. Wouldn’t it be hard to dissipate the heat generated?
    I don’t think a fan is helpful, nor any heatsink.

  11. Alas, superconducting logic gets less energy efficient as it gets warmer; The low thermal noise is exactly why it can be energy efficient.

    But, yes, you could get passively cold enough in a shadow in space.

  12. Cloud should move to the "space" above clouds. Free cooling and superior access to most of the world.

  13. These demonstrations show that AQFP logic is capable of both processing and memory operations and that we have a path toward practical adiabatic computing operating at high-clock rates while dissipating very little energy.

    Adiabatic quantum computing? They buried the lede.

  14. It's really kind of amazing it's taken anyone this long to apply superconducting research to integrated circuits.

  15. This solution is not scaleable. As many have pointed out, cooling to 4K is not easy. Requires helium and a helium refridgerator. Forget data centers…

  16. These might be useful as computers in space systems- if you compute in the shade you would have a much lower ambient temperature and easier heat management.

  17. "The devices have to be cooled 4.2 Kelvin to allow the AQFPs to go into the superconducting state." Just casually throw that little detail in at the end. No biggy.

  18. The problem with dissipating very little energy at 4.2K, is that your heat sink is at closer to 280K. So you're expending 66 watts in the refrigerator for every watt you expend in the processor.

    It better be 67 times more efficient than a room temperature processor!

    So, this is 80 times more efficient. Theoretically, it could save energy, but only if the refrigerator used was pretty close to theoretical limits.

Comments are closed.