Sign in / up

back to article IBM demos graphene circuits

Researchers at IBM are showing off a high-speed switching circuit based on graphene. In a paper presented in the journal Science today, Big Blue is taking its carbon-based transistor up another notch and making full integrated circuits. IBM showed off wafer-scale graphene transistors back in February 2010, demonstrating …

COMMENTS

House rules Send corrections
This topic is closed for new posts.
  1. Tony Barnes

    How to monetise

    Surely this was answered in the post - work out how to turn it off, or simply re-imagine how computers should work...!!

    This feels promising, then again, its been so long since the last carbon nano-tube hoo-haa, maybe I'm just getting excited with graphene instead

    1 0
  2. Mage Silver badge

    Turn it off?

    Umm. It only needs well defined states to build a Digital Computer. Not actual "off". Though power consumption might be a problem.

    An analogue computer is a Model of a system. It uses Analogue summing, log, Multiply, Square and Square Root circuits. Multiply by constant is an amplifier. Divide by constant is a pair of resistors. The number resolution is poor at between 1% to 0.1% full scale, thus roughly 7 bits to 10 bits resolution. The wiring decides the "model" or problem to be solved. Thus an analogue computer is not something for transistors that don't switch off so well, it's not a computer in the same sense at all as it has to be physically wired for a particular task. It's also inherently a "Data flow" processing device with parallelism and serial pipelines depending on the Process to be modelled.

    There have been digital logic gates made with devices that don't fully turn on.

    ECL

    http://en.wikipedia.org/wiki/Emitter-coupled_logic

    It's very power hungry, but was much faster than original TTL

    A few late 1950s computers used it.

    By the 1980s ECL type of design was mostly just used for Dividers from UHF or Microwave.

    3 0
  3. nyelvmark
    Boffin

    power consumption might be a problem

    Yes, I suspect this is THE problem. As you say, any 2 producible and discernable states can be used to implement a digital logic circuit, but if those states are 'large current flowing' and 'even larger current flowing' then you're certainly not looking at mobile applications, unless the speed advantage is so overwhelming that the net power use for a computation is comparable. It could have applications in supercomputing, and in GPUs, though.

    2 0
  4. TheOtherHobbbes

    ECL

    was still being used in computer designs in the late 80s.

    The last big-iron ECL CPU was probably the DEC VAX 9000 from 1989.

    This was just before a team at DEC designed an IC version with similar performance for a tiny fraction of the production cost and power requirement. (And there was shuffling of feet and awkward silence.)

    100GHz for a transistor switch isn't all that fast. IBM had 500GHz switching in 2008.

    Assuming the technology can be digitalised, by the time you've chained together a big pile of logic, propagation delays and switching delays add up to a big number. So it's unlikely a full-scale graphene CPU would be much faster than current silicon.

    I'd expect a multiplier like 2X or maybe 3X, but not much more than that.

    0 0
  5. John Savard

    Completely Off

    Microprocessors were even made from ECL, which achieved its speed by not turning the transistors completely off, but instead running them much as they would run for analog circuits. But CMOS supplanted it because being turned completely off, or completely on, nearly all the time saved energy, thus avoiding waste heat.

    And for a computer to have as many transistors as at least the very first Pentium chips lets it use efficient circuits for operations such as multiplication that give a huge advantage over a smaller computer even with faster transistors. So there is a huge threshold that any new faster circuit technology must reach.

    0 0
  6. Anonymous Coward
    Anonymous Coward

    How to monetise

    1. Get a vague idea of how it might work

    2. Patent it

    3. Wait for everyone else to catch up

    4. Profit!

    0 0
    1. Ru
      Reply Icon
      Black Helicopters

      Step 3 may come too late

      Plenty of stuff to do with image/video processing has been around for years and years. Problem is, only recently has hardware been good enough to do it cheaply and in real time, or near-real time.

      You should perhaps be more concerned about:

      1. Find someone else with a vague idea of how it might work

      2. Obtain the idea from them, and ensure their silence

      3. Wait ten years

      3. Patent it

      and so on as you originally posted.

      0 0
  7. etabeta
    Facepalm

    27 decibel conversion loss!

    "27 decibel conversion loss at 4GHz" This is a somewhat insane conversion loss for a mixer, 2 degrees of magnitude worse than an off-the-shelf Si or GaAs mixer; they still have a long way to go .

    0 2
  8. Rattus Rattus

    Missing units

    "between 25 and 125 degrees" - is that in a proper temperature scale, or that bizarro Fahrenheit nonsense?

    1 0
This topic is closed for new posts.

Other stories you might like