Sign in with
Sign up | Sign in

Researchers: We Make Transistors 1M Times More Efficient

By - Source: NWU | B 36 comments

Forget 10x efficiency increase. What about 1,000,000x?

New research coming out of the McCormick School at Northwestern University proposes the use of spin-transistors as a replacement for traditional CMOS transistors in computer chips. The scientists behind the project claim that electronic circuits could be up to a million times more efficient when using spintronics.

First shown at the International Symposium On Nanoscale Architectures held earlier this month in the Netherlands, the technology is described as an "entirely new logic circuit family" that uses magnetoresistive bipolar spin-transistors that are capable of performing the same functions as traditional CMOS transistors, but take advantage of the magnetic properties in electron spin to do so.

How fast and how efficient is the technology now? Don't ask. No information beyond the initial research and an associated paper "Emitter-Coupled Spin-Transistor Logic" has been disclosed. Don't expect this technology to materialize in any products anytime soon. However, in theory, it is amazing to see what may be possible in future devices.

Discuss
Ask a Category Expert

Create a new thread in the News comments forum about this subject

Example: Notebook, Android, SSD hard drive

This thread is closed for comments
Top Comments
  • 24 Hide
    Pyree , July 12, 2012 9:50 AM
    ^Correction again. Thousand is kilo (10^3), million is mega (10^6), billion is giga (10^9).
  • 21 Hide
    The_Trutherizer , July 12, 2012 9:51 AM
    1M x Efficiency would mean almost no heat generated meaning not only that a CPU would be able to operate without a fan (which would be great), but that you could stack as many layers of transistors on top of each other as you want. Also these could probably then be clocked at several hundred or thousand GHz if the timing mechanism allows for it. If this research leads to anything concrete we may soon see the first cube form processors with as much vertical real-estate as it has horizontally.

    In short if this is true then we may see single processors with as much computing power as a modern day super computer or more. I wonder how small they can make these transistors...
  • 20 Hide
    lashabane , July 12, 2012 9:14 AM
    In theory; I rule the world.

    It's only theory.

    I do like the direction this research is headed. If only theories were put into practice faster than the current 10-15+ year timeline we're currently seeing.
Other Comments
    Display all 36 comments.
  • 20 Hide
    lashabane , July 12, 2012 9:14 AM
    In theory; I rule the world.

    It's only theory.

    I do like the direction this research is headed. If only theories were put into practice faster than the current 10-15+ year timeline we're currently seeing.
  • 9 Hide
    AusVip3r , July 12, 2012 9:31 AM
    Efficiency in terms of energy usage or performance?
  • 24 Hide
    Pyree , July 12, 2012 9:50 AM
    ^Correction again. Thousand is kilo (10^3), million is mega (10^6), billion is giga (10^9).
  • 21 Hide
    The_Trutherizer , July 12, 2012 9:51 AM
    1M x Efficiency would mean almost no heat generated meaning not only that a CPU would be able to operate without a fan (which would be great), but that you could stack as many layers of transistors on top of each other as you want. Also these could probably then be clocked at several hundred or thousand GHz if the timing mechanism allows for it. If this research leads to anything concrete we may soon see the first cube form processors with as much vertical real-estate as it has horizontally.

    In short if this is true then we may see single processors with as much computing power as a modern day super computer or more. I wonder how small they can make these transistors...
  • 12 Hide
    hardcore_gamer , July 12, 2012 10:28 AM
    GeoManNerd correction, "1M times more efficient" would be 1000 times more efficient, the numeric abbreviation M stands for mega not million. What you want is either 1G (G for gigia, or 1 000 000) or just "1 million".


    You are wrong and the article is right.

    1 Mega = 1 Million

    1 Giga = 1 Billion
  • 18 Hide
    GeoMan , July 12, 2012 10:56 AM
    Quote:
    ^Correction again. Thousand is kilo (10^3), million is mega (10^6), billion is giga (10^9).


    Quote:
    You are wrong and the article is right.

    1 Mega = 1 Million

    1 Giga = 1 Billion


    I fail, that's what I get for trying to think before my first cup of coffee (-;
  • 14 Hide
    Anonymous , July 12, 2012 11:49 AM
    "an associated paper "Emitter-Coupled Spin-Transistor Logic""
    Where is this paper? Please add a link to the Paper or the Journal. I'm sick of articles without references.
  • -2 Hide
    bepi , July 12, 2012 12:04 PM
    Think about that: When a technology like this will become real, it will be very hard to produce working chips for sure. As a result we will see small CPUs with one core and a very limited functionality. The chip may look like a 8086, but if it is 1.000.000 times more efficient/fast(?) it will be competitive anyway.
  • 5 Hide
    gtvr , July 12, 2012 12:21 PM
    I don't think they have looked at mass production techniques yet. I'm certain that over time, something like this would start out slow but get more efficient as techniques are developed.
  • 4 Hide
    thecolorblue , July 12, 2012 12:26 PM
    very cool
  • 5 Hide
    Pyree , July 12, 2012 12:57 PM
    casorati"an associated paper "Emitter-Coupled Spin-Transistor Logic""Where is this paper? Please add a link to the Paper or the Journal. I'm sick of articles without references.

    It is a conference paper for NANOARCH. The NANOARCH symposium proceedings for 2012 is yet to be published online. See here:

    http://ieeexplore.ieee.org/xpl/conhome.jsp?punumber=1001687

    The paper is :

    Emitter-Coupled Spin-Transistor Logic

    J. S. Friedman, Y. I. Ismail, G. Memik, A. V. Sahakian, and B. W. Wessels

    In Proc. of IEEE/ACM International Symposium on Nanoscale Architectures (NANOARCH), Amsterdam, The Netherlands, July 2012
  • 10 Hide
    freggo , July 12, 2012 2:10 PM
    GeoManNerd correction, "1M times more efficient" would be 1000 times more efficient, the numeric abbreviation M stands for mega not million. What you want is either 1G (G for gigia, or 1 000 000) or just "1 million".



    Uh...
    k = kilo = 1000 (every drug dealer knows that)
    m = mega = 1.000.000 (Lottery winners know that)
    g = giga = 1.000.000.000 ( minimum to get a Fortune 500 listing )
    t = tera = 1.000.000.000.000 ( Government debts...)
    :) 


  • 3 Hide
    thehidecheck , July 12, 2012 2:27 PM
    and moore's law continues uninterrupted.
  • 7 Hide
    CaedenV , July 12, 2012 2:56 PM
    thehidecheckand moore's law continues uninterrupted.

    how about: and more's law is beat up in a back alley and left for dead. If this works as advertized, and even comes out in 10 years it would still be more than the mere 'doubling' effect of more's law
    freggoUh...k = kilo = 1000 (every drug dealer knows that)m = mega = 1.000.000 (Lottery winners know that)g = giga = 1.000.000.000 ( minimum to get a Fortune 500 listing )t = tera = 1.000.000.000.000 ( Government debts...)

    you forgot peta, exa, zetta, yotta, zona, weka, vunda, uda, treda, sorta, rinta, quexa, pepta, ocha, nena, minga, and luma.... not sure what comes after luma...
    http://jimvb.home.mindspring.com/unitsystem.htm
    I want a 1Lumabyte HDD... I think it would last me a while :D 
  • 0 Hide
    ashinms , July 12, 2012 3:24 PM
    The_Trutherizer1M x Efficiency would mean almost no heat generated meaning not only that a CPU would be able to operate without a fan (which would be great), but that you could stack as many layers of transistors on top of each other as you want. Also these could probably then be clocked at several hundred or thousand GHz if the timing mechanism allows for it. If this research leads to anything concrete we may soon see the first cube form processors with as much vertical real-estate as it has horizontally. In short if this is true then we may see single processors with as much computing power as a modern day super computer or more. I wonder how small they can make these transistors...


    I hate to sound cynical, but that's kinda what Moore's law is all about. We really shouldn't be so surprised. I understand there is no guarantee that Moore's law will continue on, but if my life was at stake on a bet, I would have to say that in ten years, we could be looking at petaflop desktops.
  • 2 Hide
    tolham , July 12, 2012 3:29 PM
    "spintronics"

    sounds like a lame punk band.
  • 1 Hide
    phamhlam , July 12, 2012 3:34 PM
    I just want my Graphene. Graphene everything from CPU to the coolers on top of CPU.
  • 0 Hide
    fb39ca4 , July 12, 2012 3:38 PM
    Of course it will be three-to-five years from now :( 
  • 1 Hide
    fb39ca4 , July 12, 2012 3:39 PM
    phamhlamI just want my Graphene. Graphene everything from CPU to the coolers on top of CPU.

    It doesn't work for digital electronics like CPUs, they are developed for radios.
Display more comments