Standards Wanted

Over the next few years, as the industry moves to stacked die of various sorts—2.5D, 3D, system-in-package, package-on-package, and probably some others we haven’t considered yet—there will be a major need for standards. We will need standards for placement, interconnects, power leakage, characterization of IP and a slew of other things we haven’t even thought about yet.

Most companies don’t like to talk about this stuff ahead of time. From the public statements by companies you’d be hard-pressed to even figure out who’s developing 3D stacks right now because they consider that competitive information. But what’s different this time is that the majority of systems, subsystems and various other forms of IP that will be put together in these stacks won’t be internally developed.

There will certainly be proprietary pieces. High-performance processors, low-power configurations, analog IP will all be unique and command a high value. But the whole idea behind stacking of die is that you can build other pieces into the stack that you don’t have to develop yourself.

To avoid problems with physical effects such as heat, noise, electrostatic discharge and a raft of other issues, some sort of standards need to be in place to determine where these heat-generating parts are located on a die, how close they can be to a through-silicon via, and how close you can put analog and digital parts and still have a working chip. Two known good die do not necessarily equal one good stacked die.

It’s in everyone’s interest to begin discussing this kind of information much more openly than in the past—including the big IDMs, which multiple sources tell us are all working on stacking approaches even though they look away when asked directly. We hear lots of complaints about multiple power formats (UPF 1.0, IEEE 1801 and CPF) from companies that are using those formats to verify chips. That’s just a minor annoyance compared to the problems that can turn up in a stacked die without consistent standards. It’s time to start talking and thinking about the problems that can arise—and how to avoid them as an industry.

–Ed Sperling