back to article Meet the 1,000 core chip that can be powered by an AA battery

Six years after University of Glasgow researchers first achieved the feat, an American university has demonstrated a 1,000 core processor. While Glasgow used a FPGA, the “kilocore”silicon produced by the University of California Davis’ VLSI* Computation Lab differs by putting 1,000 independently programmable cores on a single …

  1. James 51
    Joke

    So we know which CPU the iPhone 7 will be using then.

    1. Anonymous Coward
      Anonymous Coward

      So we know which CPU the iPhone 7 will be using then.

      Not unless they reduce the battery size - you can't possible produce an iPhone with a battery that lasts longer than a 6 pack of beer..

      1. Matt Bryant Silver badge
        Happy

        Re: AC

        "....battery...." And an AA battery isn't proprietary enough for Apple, they would want an iAA battery with it's own proprietary terminals that could only charge off an Apple outlet and would cost $300.

        1. Oengus

          Re: AC

          iAA battery with it's own proprietary terminals that could only charge off an Apple outlet and would cost $300."

          iAA battery with it's own Patented design proprietary terminals that could only charge off an Apple outlet and would cost the consumer $300 (but only cost Apple 30 cents).

          FTFY

          Of course if you deigned to infringe on their patent they would sue you into oblivion...

          1. james 68

            Re: AC

            Or more likely they'd just use an AA battery and yell "First!!!!" and sue everybody anyway arguing that because they yelled first that means that they obviously designed it so only they can use it (just like everything else- rounded corners included).

  2. Mage Silver badge

    13.1 W.

    Not realistically an AA cell. That's 10 amps at 1.3V.

    An Alkaline AA cell is 1.56V (fresh) to 0.9V at end point and really ought not to have more than a 50mA load for efficiency. An NiMH at 10mA might have half the capacity as Alkaline but at 1A (plausible maximum) about x5 capacity due to MUCH lower internal resistance, though probably half the label’s AH rating at such high current draw..

    Perhaps 20 paralleled Alkaline D Cells, or five paralleled NiMH D cells to have a 10 A rating.

    So for half an hour on an AA size NiMH you'd need a bit less than 1.3W load. A tenth.

    1. Swarthy
      Meh

      Re: 13.1 W.

      A tenth - like in the part of the article that mentioned the AA Battery? According to the researchers, the chip can execute 115 billion operators a second while dissipating only 0.7W, requiring only a AA battery.

      The 13.1W TDP was for 1 Trillion Ops/Sec (TrOPS?) slowing down to ~100 Billion OPS(0.10 TrOPS) lowered the power requirement to a AA battery.

      1. Mage Silver badge

        Re: 13.1 W.

        So when practically idling it might feasibly run off a D cell.

        0.7W still isn’t a realistic load, for AA, only possible.

        Still sounds like marketing hype. No doubt it's very clever, but mention of an AA cell would be more appropriate for wearable or portable gear at 100th of the normal running power.

        1. JeffyPoooh
          Pint

          Re: 13.1 W.

          Marketing hype.

          If it's such low power, then why does the picture show huge red and black (power?) cables [gauge 000?], and then also so many red and black (power?) wires and so many fat SMT capacitors?

          It sure doesn't look like an ultra low power circuit card...

          Too much BS marketing hype. Where's Henry VIII when you need him?

    2. TheVogon

      Re: 13.1 W.

      "Not realistically an AA cell. That's 10 amps at 1.3V"

      Or 3.54 Amps at 3.7 Volts as per say a Lithium AA (14500) cell.... Pushing the limits but quite possible.

      But actually it says 0.7W for an AA - so that's less than 200mA on a Lithium AA cell...

  3. jms222

    Given that there are few FPGAs that can really run off AA and do anything seriously computational there's something not quite right here. In fact more than something.

    Also look at the quantity of red and black cabling to power the thing !!

    1. John Brown (no body) Silver badge

      That's the temporary prototype video lash up so they can prove it plays Crysis!

  4. Prst. V.Jeltz Silver badge
    Joke

    Blimey thats big , is that powered by AA battery as in "Anti Aircraft Battery" ?

  5. Mephistro
    Coat

    "...the chip can execute 115 billion operators a second..."

    So this is the end for unemployment in IT in the whole Galaxy, isn't it?

    Hmmm... wait...

  6. MT Field

    Why?

    There must be a reason for building it - some application that needs this architecture - otherwise ... title?

    1. chris 17 Silver badge

      Re: Why?

      surely just because?

    2. Mellipop

      Re: Why?

      Weather forecasting, DNA sequencing, speeding up my browser. There's 3.

      Really? You can think of anything that has lots of data points that could benefit from being processed together?

    3. Alister

      Re: Why?

      There must be a reason for building it - some application that needs this architecture - otherwise ... title?

      @MT Field

      This outlook on life really annoys me. If everyone had this attitude, there would never be any innovation, no new discoveries, no experimentation.

      I hope you will choose not to take advantage of this or any other technical innovation, in fact, I think you should stop using your mobile phone, laptop, PC or whatever it is you use to post on here, and go back to rubbing two sticks together for warmth.

    4. Ben Tasker

      Re: Why?

      It's part of Skynet:

      ....the chip can execute 115 billion operators a second while....

      But seriously, as others have said - does there need to be a "why" for trying everything new? Once a technology is developed, uses will generally be found for it, and otherwise unthought of technologies sometimes grow up around them

      1. Prst. V.Jeltz Silver badge

        Re: Why?

        i agree with the above replies , but surely 4k is enough pixels for a tv?

  7. Rob Booth

    I'm no supercomputer, but isn't 1000 less than 4096?

    http://www.extremetech.com/extreme/187612-ibm-cracks-open-a-new-era-of-computing-with-brain-like-chip-4096-cores-1-million-neurons-5-4-billion-transistors

  8. This post has been deleted by its author

  9. MatsSvensson

    How is the work on the inertial shock dampers progressing?

  10. Michael Hoffmann Silver badge
    Joke

    In the case of collapse of civilisation...

    ... can it be powered by a potato?

  11. Bob H

    Reminds me of the Epiphany-IV 64-core Microprocessor by Adapteva, sure it has less cores but the architecture seems similar. The problem with the Adapteva and I imagine a similar problem for this 1000 core design, the on-core memory is tiny and so it is difficult to fit useful workloads in to them. So you spend loads of time on an external CPU with a bigger core scheduling the tasks to the tiny cores. The architecture is really hard to programme for.

    http://www.adapteva.com/products/e64g401/

  12. Mellipop

    missing and broken links

    Mr. Orlowski mentions Glasgow (university?) Doing this 6 years ago, but no link and no mention on UC Davis list here - http://vcl.ece.ucdavis.edu/misc/many-core.html

    Actually Green Arrays 144 core chip also isn't mentioned on that list. Probably because it's designed by Charles (Chuck) Moore and is programmed in Forth - http://www.greenarraychips.com/

    If anyone can help with the last link in the article about awards that would be interesting. Ta.

    1. Uffish

      Re: missing and broken links

      Try typing in "Glasgow university 1000 core fpga" into Startpage. It was the first result when I tried it.

  13. Ogi

    Sounds like a transputer

    I remember these: https://en.wikipedia.org/wiki/Transputer

    Essentially the idea was to link up many processors using an interconnect, in order to provide stupid amounts of parallelism.

    It didn't take off at the time because CPU improvements in clock speed and die size was surpassing Inmos'es ability to keep up, so they were always expensive for what they were. Also people by and large were still programming for single core machines, and there was a lack of tools and compilers for it. As such getting programmers for it was expensive too.

    However now as we are hitting a slowdown in die shrinking, along with a general improvement in tools/compilers/languages for programming parallel systems, perhaps it is time to revisit transputers. I have been thinking it is since GPGPUs started taking off.

    1. Anonymous Coward
      Anonymous Coward

      Re: Sounds like a transputer

      there was a lack of tools and compilers for it

      I can recall when this emerged - that's REALLY a long time back, like the mid 1980s. AFAIK, the only language available to program for it was OCCAM.

      It would indeed be interesting to revisit that.

      1. AceRimmer1980
        Pint

        Re: Sounds like a transputer

        I used to work at a small company in Bristol which made image processing boards based around Transputers. The most common model was the T800 <obvious 'hasta la vista' joke />

        As Occam was as much fun as chewing your own limbs, we were also provided with a Parallel-C compiler, which would scale the code over however many cores were available. It was pretty funny to see these boards in a £1000 PC, the cost of which was dwarfed by these boards, and was only being used for PSU/file storage anyway :-D

        As the earlier correspondent noted, they were a neat idea which got overtaken by ChipZilla's R&D budget. But in this day and age, some of the run-time paradigms and algorithms might return to relevance.

        Pint for my interview for this job taking place in the nearby pub.

    2. Vic

      Re: Sounds like a transputer

      It didn't take off at the time

      It did take off at the time - many units were sold.

      And even after the Transputer name ceased to be used, the cores were still selling - as the ST20, which powers a significant portion[1] of the digital TV decoder market.

      Disclosure: I used to work there...

      Vic.

      [1] At one point, the ST20 had almost 100% of the European decoder market. But then the competition came along :-)

  14. Mike 16

    Maybe a single chip CM-1?

    So maybe the application is sequencing Dinosaur DNA so biologists who slept through their high-school biology class and so don't know about gender assignment in amphibians can get funding from an amusement park.

  15. John Savard

    Better than the Press Release!

    I had heard about this processor earlier from news articles that were basically just the press release. Kudos to The Register for providing some real data on this chip.

    Apparently, the individual chips are built to work on 16-bit data, and they have a miniscule amount of internal memory. No cache, and in-order. So this is not a design that could be immediately put into the next big supercomputer to increase its performance; it would be next to useless in such an application.

    But it is still important. If it succeeds, and shows that one can get 1,000 processors working effectively on a single chip, then that part of the technology can be applied to a die with 2 billion transistors instead of 600 million, on a newer process, with the individual processors being big enough to be effective when applied to problems of interest.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon