back to article IBM's $3bn bet on next-gen computers: Carbon nanotubes, neuro chips

IBM boffins have been given a tidy $3bn cash pile to solve a problem that lurks not too far in our future. That problem is the imminent breakdown in conventional chip operation and chip materials as we shrink transistor gates from today's 14nm process size to 10nm and 7nm. At around 7nm, which most industry observers expect …

  1. Terry Cloth
    FAIL

    And how about end-of-life?

    GaAs? You think it's tough recycling your electronics now?

    And what are the breakdown modes (if any) of graphene and carbon nanotubes? At a time when we're finding plastic slivers in animals living in the abyss [1] around volcanic hot spots, it would be nice to see us determining how to safely dispose of new materials before we start integrating them into our environment.

    ------

    [1] http://www.vliz.be/imisdocs/publications/251718.pdf

    1. Destroy All Monsters Silver badge
      Holmes

      Re: And how about end-of-life?

      I don't think arsenic or gallium as dopants will ever be a large problem.

      Indeed, naturally occuring arsenic in groundwater is a problem NOW.

  2. Frumious Bandersnatch

    how about this for a solution?

    A quite significant amount of space has to be devoted to lines for carrying the clock signal. Instead of etching lines to carry this signal on the silicon, what about using radio emissions in selected bands to keep individual parts in sync with each other? In fact, why stop there? Tiny directional antennae would give near-perfect fanout and it might open up new parallel processing capabilities. Throw in a few diffraction gratings (created by etching regular logic areas on the silicon) and you could claim to have some quantum-level processing available too.

    Of course, I know next to nothing about these things, so this idea is almost definitely a crock of shit.

    Anyway, for the real reason I wanted to post: thumbs up for "scrying". Take that, spell-checker!

    1. Destroy All Monsters Silver badge

      Re: how about this for a solution?

      Given the size of the antennas needed, these "radio" emissions would be more like "X-ray" emissions. Tough to emit and receive without big machinery and collateral damage.

    2. Anonymous Coward
      Anonymous Coward

      Re: how about this for a solution?

      Seems "easier" making chips asynchronous, and avoiding the problems of global clock distribution entirely.

      1. John Smith 19 Gold badge
        Joke

        Re: how about this for a solution?

        "Seems "easier" making chips asynchronous, and avoiding the problems of global clock distribution entirely."

        Heresy.

        Clockless logic is the work of the Devil and threatens profits. How could Intel expect to make those kind of bonuses when all chips run at their fastest?

        Burn the unbeliever.

        1. Anonymous Coward
          Anonymous Coward

          Re: how about this for a solution?

          Just because a chip CAN run faster doesn't mean it must be allowed to do so. They could continue to segment them with their random incomprehensible product numbers, and limit the throttle on the lower numbered ones.

          They'd also be able to continue to segment by power usage, which is more important than performance to most people these days (since modern quad core CPUs are more than fast enough for the vast majority of common PC tasks)

  3. John Savard Silver badge

    Well

    It wouldn't be a bad thing if, when silicon ran out of steam, Intel was left in the lurch, and IBM - chastened by its old antitrust suits - roared back.

    It would be inconvenient if Microsoft Windows had to jump from the x86 architecture to, say, the PowerPC, the way the Macintosh jumped twice, from the 680x0 to the PowerPC and then from that to the x86, but Intel's monopoly is a bad thing.

  4. Don Jefe

    Genius!

    It's uncanny, the serendipitous coincidence is certainly the work of decades of research into improbability theory. To think, the $2.3 billion pending sale of a server division to an overseas company happening in the same fiscal year as the announcement of a $3 billion research initiative that also meets penalty and tax free fund repatriation regulations for R&D spend.

    The odds of that all happening at once are unfathomable. If often seems fortunate when so many great things just happen to fall into place at the same time, but you've got to be careful. You can quickly find yourself in a negative luck equity situation. I sure hope IBM doesn't do something crazy now and enter into long term funds matching projects with the Federal government. Surely they can't be so lucky as to get out of taxes on repatriated funds and have the taxpayer cover several billion in IP generating research that the government will buy in the future. Can any one company be so very lucky?

    1. Destroy All Monsters Silver badge
      Holmes

      Re: Genius!

      Someone seems to be a bit miffed that the extremely well-going F-35 forever-development might not be fully funded for another decade or two at the same time as the Obamacare ensure-your-liver-chirrosis hoi polloi voting ploy.

      On the other hand, I'm sure tax rebates (if any) will not accrue fully at the start of an announcement about a "research project" that is so far mainly fluff for a Wall Street embiggening.

      1. Don Jefe

        Re: Genius!

        My liver uses cutting edge sugar reclamation technology that keeps it supersaturated in pure grain alcohol which acts as a preservative. Half measures are deadly you know.

        R&D funds repatriation regulations aren't a rebate, but an exemption. Those regulations are why I call bullshit when large companies cite taxation as an impediment to R&D. You're allowed to completely break it all out; for example, instead of booking $1B in stateside funds as R&D you can book $1B in offshore funds as R&D and that $1B is not taxed. It's a sweet deal, the liberation of existing stateside funds means you now have an $1B in tax exempt (the taxes were already paid on that money) funds to allocate as you wish. Modify departmental margins, offset debt service costs, buyback stock, use as a one time dividend or throw a really big party for yourself.

        It's all another of those taxation mechanisms that is perfectly legal, but you've got to be shifting really large amounts of money for it to be worthwhile. But it's also one of those things executives, bent politicians and idiots in the general public 'forget' to mention when raging about taxes. There is, and has been for ~20 years, a direct, dollar for dollar reward for companies willing to actually invest in R&D. A multifaceted reward that affects every aspect of a publicly traded company in the US, but companies don't actually want to do R&D. They just don't want to pay taxes.

  5. itsallgonepearshaped

    Coincidence?

    That IBM is researching the next phase of CPU and at the same time exiting the existing the current phase? I'm guessing it's not going to be compatible, so maybe we're only looking at a few years before the next revolution?

  6. Martin Budden Silver badge

    Meanwhile, the whole tech industry is built around the promise that chips regularly get faster and more capable. Unless someone solves the post-7nm challenge, the IT wagon could jump the tracks.

    It might actually be a good thing if the focus has to shift to trimming the fat from bloatware.

  7. John Smith 19 Gold badge
    Holmes

    "as we see it is coming to a point where a lot of new innovation is needed"

    Handy hint.

    When you can measure gate widths in the number of atoms across that gate as no more than a 3 digit number the end is approaching.

    If you're talking about density of data your options are pretty limited.

    Shrink the gate width.

    Well that's already happening.

    Shrink the thickness of the chip (and lay several of them on top of each other)

    Starting to happen but that IMHO can go much further. But if you think the heat output of current processors is fierce that's just getting started.

    "Quantum" computing (how many definitions of this are there?). If 1 atom can have many quantum states then it's carrying out the functions of hardware made up of many more atoms. Can the number of Q states the atom can take up (and therefor "replace" other atoms in a more normal structure by doing so) make up for the huge number of atoms needed in all the peripheral gubbins needed to read and write it? Can you come up with I/O devices made up of just a few atoms themselves?

    And of course there's my personal favorite in the tinfoil hat SEL stakes "engineering" the energy level transitions of individual atoms.

    No I don't know what that means, but it sounds cool as f**k.

  8. knarf

    there's a reason why semiconductor engineers like beer so much.

    Because it tastes sooooooooooo nice and makes coding easier, well it did when I was a student, a jnr, a dev and a lead.

  9. Anonymous Coward
    Anonymous Coward

    "Meanwhile, the whole tech industry is built around the promise that chips regularly get faster and more capable. Unless someone solves the post-7nm challenge, the IT wagon could jump the tracks."

    we need faster chips? no, we need less crud on web sites.

  10. Battlescarred

    O tempora, o mores...

    "Basically, silicon technology as we see it is coming to a point where a lot of new innovation is needed," IBM Research's director of Physical Sciences Supratik Guha told The Register. (I quote). "New innovation"??? When was innovation ever not new?! Still, IBM is the organisation that invented that dreadful solecism on the labels of magnetic tape used for its software: "This media contains copyright material", which of course should have read 'this medium'. But that's another lost cause, along with 'outside of'', 'very unique' and other horrors that even the BBC is now perpetrating.

    1. Frumious Bandersnatch

      Re: O tempora, o mores...

      re: "outside of"

      People having been using that for quite a while, viz.:

      "Outside of a dog, a book is a man's best friend. Inside of a dog, it's too dark to read" --- Groucho Marx

      1. Don Jefe

        Re: O tempora, o mores...

        'Very unique' is kind of like that too. The 'very' part can be a differentiator between one of a group of similar, but not identical things, like dogs. 'I use Bloodhounds as Prole Hounds on my estate. The pack leader is very unique because I had fully functional Human arms grafted to it because it was cheaper than engineering paw friendly firearms'. Before that Lord of the Seas Quetzelquatiol Fett Esq. was nearly indistinguishable from his pack mates.

  11. Anonymous Coward
    Anonymous Coward

    IBMs response to the Machine announcement by HP. More power (literally) to them, looks like HP's headed the other direction to lower power and newer architectures. I can't cool the servers I have now, I'm unlikely to look for hotter ones in the fu-ture-ture-ture...

  12. Anonymous Coward
    Anonymous Coward

    Neuromorphic

    Just try to get the exact same result twice in a row. And if you solve the problem by propagating a static set of states, then just try to patch the mess and have the newly patched mush give the same answer that the unpatched mush had. If the domains are limited in a feeble attempt to solve THAT problem, then the performance is reduced sometimes to the point that the endeavor is a loss even with a fleet of propeller heads doing the work. (genetic code generators are enough of a PITA thank you)

    Banks really hate inconsistent results, although it may be good enough for government work.

    Been there, done that. I'd like to be a paid member of the team, since I can do whatever I want and not affect the outcome of the project.

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

Biting the hand that feeds IT © 1998–2021