back to article Moore's Law savior EUV faces uncertain future

The optical lithography that etches the chips in your digital devices is reaching its limits, but exactly when its oft-touted replacement – extreme ultraviolet lithography, commonly known as EUV – will be ready for prime time remains unclear. "There are still some technical challenges which, of course, lead to a certain degree …

COMMENTS

This topic is closed for new posts.
  1. Christian Berger

    EBL sounds interesting

    ...mostly because it can scale down. You can build a much cheaper lithography setup at a much lower cost. This would lower the entry point into semiconductor manufacturing. I personally believe that once chip manufacturing reaches the realm of a small company or even hackerspace, there will be new waves of innovation. Just like the home computer revolution did.

    1. Anonymous Coward
      Anonymous Coward

      Re: EBL sounds interesting

      Unfortunately for the wannabe DIY chipbuilder there's more than just lithography between a wafer of blank silicon and a fully functioning connectable chip. I can't see many of those process stages being DIY operations.

      1. Morg

        Re: EBL sounds interesting

        Not quite, the rest of the process stages are much much much easier ;)

        But still, even with EBL you'll need a clean room etc. - and that makes it much less likely.

        What is instead very likely is that as chip etching becomes more and more available, you will be able to have your own CPU designs created (although not at the top node, and not in 1-unit quantities), just like you can now have any CAD design be made in almost any material online.

    2. Anonymous Coward
      Anonymous Coward

      Re: EBL sounds interesting

      Well as the top of the range Xeon is knocking on the door of 2 billion processors, with even a low end chip knocking around 45 million, that's going to be one hell of a home brew chip you'll need to knock up.

      1. Morg

        Re: EBL sounds interesting

        First of all it's transistors not processors, and second the Xeon was not designed transistor by transistor.

        Many transistors are just cache (and that's just how close to copy/paste you can get).

        The bigger part of chip design today is assembling blocks OR modifying blocks. but it's really mostly about assembling, there isn't much transistor-level design in today's chips.

        Of course those chips remain incredibly complex and shit, but the 2 billion transistor number is a *Retarded* point of view.

        You are talking about an 8 core chip, so you might as well cut those 2 billion in 8, down to 250 million, drop the cache (prolly around 1/4th ) you're down to 180 million, and then you can start cutting that into blocks (and no I don't think the FPU change *a lot* between cpu generations ;), same can be said for the x86, SSE, avx instruction sets, arithmetic and .. well most of the chip's building blocks anyway.

  2. auburnman

    While it's good to see we're breaking through another technological barrier, part of me wants to see hardware advancement stagnate for a brief period in the hope of inspiring companies to focus on efficient software in order to stand out.

  3. Stephen Booth

    EBL the 3D printer of chip manufacture

    "E-beam lithography itself is pretty straightforward," Liebmann said. "We all sort of qualitatively understand how this all works. And we all also understand that the biggest problem is throughput. What all of these systems are working on is massively parallelizing the system to get to a point where you can begin to make this profitable."

    EBL may not have the throughput for profitable mass production but as I understand it you don't have to make masks either so the cost should be similar for small volumes and prototyping as it is for large volumes.

    Just as 3D printers (much more expensive than mass produced injection mouldings) open up new possibilities EBL might already be interesting for the low volume specialist designs that currently use FPGAs. Maybe what we will see is the current small number of dominant chip designs built in huge global fabs being replaced by a wide range of custom designs printed to order. This might make things profitable even if there is a big hiccup in the expected price reduction with each processor generation.

    1. Morg

      Re: EBL the 3D printer of chip manufacture

      Well . not quite.

      The thing is yes you can *easily* etch just about anything, but the handling is way more complicated if there are different sizes different chips etc. on every wafer.

      Besides, there's always a percentage of bad dies, so if you just printed it once and it doesn't work you're screwed -> i.e. you need to have a minimum order (make that one wafer for example) - still a lot of overhead due to number of different designs / etc.

      You have to take into account that designs are generally reviewed several times before the final printing, and tweaking of the manufacturing process is often required to reach the desired yields. Etc.

      All in all, yes you may one day be able to print any chip you want, but you will never have access to the last 5 nodes in such a reliable manner as you described (i.E. print it - it works).

      This means that even if your chip was 3.7 times more efficient than a mainstream chip, it'd end up just as bad because of the scaling factors ;)

      There will be major changes in the silicon industry, and there will be many more players (as today shows) involved than just good old IBM / Intel / AMD (and a few minor others) - as we can see today : many chinese developing different CPU's / APU's etc. including a lot of ARM (outside of china as well), some Alpha etc.

      BUT . the reason Intel is alive today is because they were first on almost every single node . As long as research and deployment of fabs remains *that* expensive, the number of market players will remain small.

      We see it today, where medfield can hold it's own against ARM just because it's one and a half node later (32nm vs 45nm)

      We'll see it again tomorrow, when Intel releases Haswell with a crappy oversized IGP, one node below Trinity and just about as power efficient / capable. - another fab win for fabzilla ;)

      1. Anonymous Coward
        Anonymous Coward

        Re: EBL the 3D printer of chip manufacture

        "there's always a percentage of bad dies"

        True.

        Now, hands up who remembers "wafer scale integration" from the 1970s.

        Hands up who recognises the name Ivor Catt (stop sniggering, it's the best two British kids TV programs EVER, in one name). If you don't, go reading.

        Now, hands up who thinks WSI will be back? Or who has any better ideas?

  4. Alan Firminger

    Some of EUV's complexities are due to how its light is generated.

    But it does not use light.

  5. blondie101
    Happy

    one manufactoror is still confident

    ASML expect to ship production EUV in 2013: http://www.asml.com/asml/show.do?ctx=41905&rid=41906

  6. Bill Neal
    WTF?

    Nikon?

    Who makes these machines anyway? I mean the various chip printing machines. Who manufactures them? Does intel make their own, or is there some competitive market that we never hear about?

    1. blondie101
      Holmes

      Re: Nikon?

      Yep Nikon is a big name, but the biggest is ASML, a Dutch company.

      1. Anonymous Coward
        Anonymous Coward

        Re: Nikon?

        Why haven't Intel just bought ASML? Presumably Intel don't permit ASML to sell the leading edge stuff to anyone else anyway?

  7. John Smith 19 Gold badge
    Meh

    so EUV == X-Ray Projection Lithography. The designation given the wavelength looked odd.

    And now I know why. I guess changing the name did not make the technology *any* easier to make work.

    I recall Electronics Design articles talking about laser driven metal plasma soft x-ray sources in the 80's. People were saying the load/lock procedures to retain the UHV were a drag on cycle time and the resists exposure times were either *very* long or you needed to put multiple levels on.

    I doubt *anyone* thought they would still be *trying* to get it to work about 30 years later.

    Anyone still trying with storage rings?

This topic is closed for new posts.

Other stories you might like