back to article Hand in hand, TSMC, ARM head to 7nm server chip land

ARM and TSMC today announced they are working together to make chips featuring 7nm FinFETs a reality. This follows on from their work on 16nm and 10nm FinFETs. Taiwan-based TSMC describes itself as "the world’s largest dedicated semiconductor foundry," and it churns out chips for the likes of Nvidia, AMD, Qualcomm, Apple, …

  1. Alan Brown Silver badge

    understatement

    "It's extremely tricky using 193nm UV light to construct transistor gates smaller than 22nm, let alone 7nm"

    20 years ago the Great White Hope was x-ray lithography. I guess it could still be on the back burner.

    1. Jan 0

      Re: understatement

      X-ray lithography has been renamed EUV. (Is this to avoid scaring the uneducated, like renaming NMR to MRI?)

      There is another great white hope: multi electron beam lithoghraphy.

      1. imanidiot Silver badge

        Re: understatement

        Mostly because actual Xray photons (defined as between 10 nm and 0.01 nm wavelength) were found to knock loose too many electrons in the resist material and cause fuzzy edges, negating the minimum feature size gain. This is still a problem with EUV (13,5 nm wavelength) but to a lesser extend.

  2. John Sager

    Key Stage 2

    Is this now the target audience? Or are we now to assume that the average software guy has zero exposure to basic electronics?

    KS2 in UK is around 4th or 5th Grade in US, I think.

    1. diodesign Silver badge

      Re: Key Stage 2

      I either write too technical or too simple, judging from comments. With millions of readers, it is impossible to please everyone. If you know how transistors work, you don't need my article :-)

      If you're skimming El Reg for news and can't remember or don't know how semiconductors work, I hope this has helped.

      C.

  3. DougS Silver badge

    Process naming

    There isn't a standard for what constitutes an "x" nanometer process. It is likely that Intel's 10nm will be pretty similar to TSMC's 7nm so if they are out of the gate at roughly the same time then all it will mean is that the foundries have finally caught Intel (thanks to all the smartphone SoC revenue) and not surpassed them.

    1. Rik Myslewski

      Re: Process naming

      " ... out of the gate at roughly the same time" — ah, transistor humor ...

  4. Doctor Syntax Silver badge

    "You have a button on top, called the gate"

    Tricky little blighters to push.

  5. Pseu Donyme

    Somehow this reminded me of the boast: "In the USSR we have the biggest microchips in the world." (a mock one making fun of their propaganda, of course)

    1. billse10
      Coat

      http://micro.magnet.fsu.edu/creatures/pages/russians.html

      "VAX - when you care enough to steal the very best "

      :-)

  6. Anonymous Coward
    Anonymous Coward

    Getting to the limit

    Silicon has an atomic radius empirically of 111 pm, so this makes a 7 nm track 63 atoms in width.

  7. imanidiot Silver badge
    Boffin

    EUV is certainly not dead

    EUV is coming. Development is further along than you seem to imply here. Yes, it has taken much longer than expected but when you are fighting the current limits of many many fields of physics and chemistry things don't always go as planned.

    Many, many new techniques and processes have been developed just to make EUV a reality. Working tools now exist and are being used for development.

    I work in this particular field. If I look over my left shoulder I can see 2 EUV sources being assembled right now in fact.

    1. Rik Myslewski

      Re: EUV is certainly not dead

      @Imanidiot: Interesting — what's the wafer/hour throughput of the devices you mentioned? What's the wafer size? What's the power consumption? Just curious, really.

      1. imanidiot Silver badge

        Re: EUV is certainly not dead

        I know its been a while but maybe you still read this. Throughput is currently a company secret but lets say sufficient for proto run production, with full production speeds in the pipeline very soon. Wafer size is standard 300mm (noone is planning a 450 fab so no point in building a machine for 450mm. Logistics alone at that size become a challenge). Powerconsumption is again a company secret but relatively high for the actual EUV light power actually reaching the wafers. (kWatts in hundred something Watts out.)

        Sorry I can't be more specific. NDAs are a pain.

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 © 1998–2019