Materials, Architectures And Gordon Moore

Shrinking feature sizes is so 45nm. The future is much more interesting—and challenging.


Shrinking features on bulk CMOS using planar transistors has turned the semiconductor industry from a startup industry to one of the most efficient and robust industries in the world. Each new process node increases the number of chips that can be cut out of a single wafer, literally defining economies of scale.

Gordon Moore defined the direction, which certainly created a long list of challenges to be solved, but from an engineering standpoint this wasn’t exactly new and riveting stuff. It was evolutionary, within very tight parameters, and it turned into a fully baked supply chain with clear demarcations about was on top of the supply chain and who was on the bottom.

But with 2.5D-IC designs in the pipeline, work underway with 3D-ICs, announcements of finFET manufacturing capability now beginning to spew out of the manufacturers, and more interest in silicon on insulator and other wafer substrates, the industry’s finely-stitched structure is starting to come apart at the seams.

The reason is that it’s too hard to continue shrinking features. The wires are too long, too thin, and it takes too much energy to drive signals. It’s also too expensive to create analog IP at advanced nodes. New memories will be needed along with new materials, and new manufacturing techniques will likely be required if EUV lithography misses another couple of deadlines.

To continue doing this on planar transistors using the same methods requires more engineering, more time, and a lot more money. That means far fewer players, which is why we’re seeing unusual investments being made by the giants, and the rest of the industry is focusing on moving even further away from the IDM model toward a more tightly integrated ecosystem.

All of this ripples down (or up) to the IC design world. And it shows, because things have suddenly become much more exciting. The ability to really drive up performance with dramatic improvements in energy efficiency were always possible, but what’s changed is that they’re now often the less-expensive option for many applications. Even better, they’re now an integral part of the OEMs’ marketing plans, because power and performance are key attributes for consumers.

The interesting stuff is now interesting to more people, and the results that are possible with this shift—as well as the amount of experimentation that needs to happen to propel this forward—should create a whole new wave of innovation. It’s good to be an engineer—again.