Home
OPINION

The End Is Near

As 2013 spirals to a close, what’s changed so far and how will that play out in 2014?

popularity

Looking back is easier than looking forward, and looking narrow is easier than looking wide. In 2013, there were several fundamental changes.

Change No. 1: IP is now a lucrative market. From Synopsys’ standpoint, it’s been a lucrative market for some time. But the acquisitions made by Cadence, beginning in late 2012, coupled with the push by ARM into the micro-server market and the flailing response by Intel in the micro-server and foundry markets, are just the beginning. As chips get harder to design and build, and there is simply more stuff in them, it doesn’t pay to develop everything yourself.

Change No. 2: Moore’s Law increasingly looks like a historical relic (at least for most companies). That doesn’t mean progress will end, but economics are spurring new interest in improvements at 28nm (FD-SOI, body biasing), as well as a big push toward 2.5D—particularly by TSMC and GlobalFoundries—means that not everything will have to be done at the latest process technology. There may still be a role for mixing and matching, but the needle no longer points just toward single-digit line widths.

Change No. 3: The Internet of Things seemed to have sprouted to life in 2013. The concept has been around at least since the late 1990s, but there are suddenly more things talking to things, starting with cars and cell phones and smart gadgets in the home. The first step you’ll notice is the proliferation of automatic updates. We are at the crawling stage here, and it remains to be seen just how quickly this entire phenomena ramps up. But this is still the beginning of a very big shift with enormous consequences.

So what do these changes mean? Each of them carries broad implications for every facet of the semiconductor industry, and in some cases well beyond.

Prediction No. 1: Third-party IP will become a central part of every design. In fact, given the ramp up and interest in third-party IP, it will become the basis of competition between chipmakers. Being able to tweak even standard IP and quickly integrate it is the next big challenge. Standards efforts will push in this direction, but the bigger effort may be a focus on putting more IP together up front, whether that is into subsystems, entire systems, reference designs, or even full chips.

The big opportunity here, though, is IP that includes both hardware and software that has been tightly integrated for efficiency and performance. That leads to some very interesting questions about the next steps for hardware and software IP vendors, as well as what will make IP vendors successful? Will it be ecosystems, tools, or market dynamics such as stacked die that change the playing field?

Prediction No. 2: Moore’s Law will continue to languish, but not for the obvious reasons. Lithography and finFETs are just contributors to the shift. The reality is that it’s taking longer for even the most die-hard adherents to shrinking features to move from one node to the next. Just because the technology is ready doesn’t mean companies will jump every 18 or 24 months. Time between nodes is taking longer—in the case of 16/14nm, as much as four years between introduction and production-level adoption—and that’s with a proven 20nm process that isn’t changing. At 10nm we are looking at the introduction of III-V materials to improve electron mobility, as well as implanted materials on the sides of fins, potentially new substrates, massive shifts in equipment needed to manufacture chips, and enormous costs being borne by foundries.

It’s no wonder that stacked die are suddenly looking more attractive. It’s much easier to build regular shapes on advanced nodes using repeatable processes than trying to shrink everything onto an SoC using the latest incredibly complex processes in a short market window. Any takers for 7nm mixed-signal designs?

Prediction No. 3: The Internet of Things will create a security nightmare, which will open up vast new opportunities in semiconductor engineering. Security has always been based on limited access. Think of a bank vault. And then think about things talking to things that may be located anywhere in the world using multiple wired and wireless protocols.

What happens if your car downloads the wrong software, or if someone gains access to your home network through your smart appliances talking to other things? And why didn’t Dick Cheney allow updates to his pacemaker? There are important questions that need to be asked the supply chain. That’s a first step. The second step is to provide suitable answers to them.



Leave a Reply