Cyclical Behavior

As we head into yet another DAC, this may seem like a bad case of déjà vu. While there’s plenty of new stuff—and even some new companies—there are some key trends that seem very familiar.

Here are four to consider:

1. Cloud computing. For anyone who’s old enough to remember time-sharing on university computers, cloud computing should look very familiar. The scheduling is more sophisticated, the computers are more powerful and the partitioning is more secure, but the idea is basically the same.

IBM actually invented virtualization in the late 1960s as a step forward in time sharing, allowing multiple jobs to run on the same computer at the same time. Now, it seems, companies are re-inventing the concept. Why own when you can lease?

2. Stacked die. The real 3D ICs are new, but 2.5D system-in-package isn’t. It’s the return of the multichip module, which everyone said was too slow. As it turns out, shrinkage has extended the signal paths so far that it’s now actually quicker to run some signals off chip using a wider I/O channel.

So much for the all-in-one chip. Plus, you can start attaching all sorts of new things to the core logic and memory, including everything from analog sensors to MEMS.

3. Mixed-signal designs. That leads to point number three. In the early part of the Millennium, analog companies said they had decided not to mix analog and digital on the same die because the noise from digital components would impact analog signal integrity. That problem was solved, analog engineers screamed about trying to keep up with Moore’s Law, and now they’re back off the road map.

In 2.5D and 3D stacks, analog will once again be isolated onto a separate die from digital. Ironically, the thinner die may also make analog signals more susceptible to digital noise again.

4. Reaggregation. The complexity of chips at advanced nodes is almost mind-numbing. There are multiple voltage islands, multiple modes, multiple cores, multiple I/Os, hundreds of millions of gates and just about every physical effect you can possibly imagine. In fact, there is so much complexity that there is growing speculation about certain pieces of design being reaggregated.

Acquisitions are certainly pointing in that direction. EDA has always been on an acquisition path, and buyouts are considered an acceptable exit strategy for many startups that cannot get enough momentum to go public. But what the largest players are buying up now are companies that are well outside the normal EDA market. They’re actually creating their own technology stacks, and that’s a whole different way of looking at the market.

–Ed Sperling