Siliconcowboy's Blog

Technology, journalism, social media and social responsibility

Semiconductors: End of the Drive

The notion of a purely physical continuation of Moore’s Law has to be smashed. My guess is that by 2015 – perhaps as early as 2013 – it’s essentially over.

Sure, there will be innovations to semiconductor manufacturing – in particular, new lithography systems – that will enable sub-20 nanometer production.

But we’re never again going to attain the manufacturing throughput or yields necessary to enable the doubling of density, doubling of performance, or unit cost savings that have been possible up to this point.

Whether your money is on direct-write laser systems, double patterned masks and immersion lithography, or more advanced, micro-machined systems that can line up individual atoms –  all will require more time per chip to complete their work. All will still have to deal with significant physical effects, including thermal damage, shape-based defects and cross-chip electromigration.

If you’re betting on graphene-based chips, well that’s an entirely new paradigm that will require a complete re-write of simulation, design, verification, manufacturing and test. In other words, it won’t be ready at least until the middle of the decade, if not much longer. Even then, a sub-20 nanometer manufacturing regime might be too aggressive at the outset.

If, like Intel, you are experimenting in laser-based processing, you’ve probably found one of the few innovations that will support physical scaling over the long term.

The bottom line is this: Every chip built after 2015, in the sub-20 nanometer realm, will have to be strategically critical.

These chips are likely to be hugely expensive, due either to the bidding price to get on the rare foundry line, or the cost of ramping your own manufacturing facility.

There will be fewer of these chips, so their price relative to the whole system cost will be disproportionately high.

Most important, the number of players who will succeed in getting their chips on these lines will be extremely small.

So what happens to everyone else?

Let me put it to you this way. Have you ever seen what happens when you try to drive 1,000 head of cattle through an 8-foot gate?

Right now, all of our conventional semiconductor manufacturing facilities worldwide are running at full capacity. And while industry analysts are expecting a slowdown, in part due to seasonal effects, in part because the iPad seems to be impacting notebook sales, so far there’s no real indication that things are in fact slowing. Rather, there are more gadgets – including more iPad alternatives – being built every day.

So the conventional manufacturing lines will have to expand, pretty quickly.

New paradigms in processing, packaging and design are also going to have to emerge at top speed.  Large multiprocessor designs with multiple power domains will allow chip makers to mimic the functionality of entire systems without moving to sub-20 nanometer silicon. Various forms of die stacking technology will enable three-dimensional chip design. Embedded parallel processing systems will allow the integration of serial and parallel functionality, such as chips with CPUs and GPUs on board. And HP Labs’ memristor technology will have to be exercised to its full potential.

So the net is that the semiconductor industry will continue to thrive, but it’s going to grow in fundamentally new directions. There will be sub-20nm chips, but not for everyone.  The remainder will have to innovate in other, novel ways.

Let me say that again: semiconductor innovation will have to expand in new directions, open new doors, explore new ideas.

But any good cowboy could have told you that.

After all, if you don’t want to be out there chasing doggies until well after sundown, you gotta open some more gates.

2 comments on “Semiconductors: End of the Drive

  1. Lou Covey
    October 5, 2010

    I think you’re right with the industry the way it is today, but I know technology exists now that can drop the cost and increase the throughput of new chip designs, but the industry just won’t pay for it. A large US semiconductor company has driven a technology innovator into bankruptcy over the past two years internally debating whether they can make the same thing in house. In the meantime, they’ve lost $2 billion and put the startup out of business. You can’t sustain a business with that lack of foresight. That’s the real problem.

    • siliconcowboy
      October 5, 2010

      Thanks for your response Lou. A lot of people don’t understand that in order for Moore’s Law to work, not only does the density of the semiconductor chip have to double, but the throughput and yields of the manufacturing facility must double as well. Otherwise, you get diminishing returns for your investment and the cost to everyone skyrockets.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s


This entry was posted on September 30, 2010 by in Chip+System Design, Tech Trends.
%d bloggers like this: