Spencer Reiss talked with Mead, who turned 70 this year, at his house among the redwoods in Woodside, CA.

Technology Review: You’re famous for saying, “Listen to the technology.”

CARVER MEAD: To understand reality, you have to understand how things work. If you do that, you can start to do engineering with it, build things. And if you can’t, whatever you’re doing probably isn’t good science. To me, engineering and science aren’t separate endeavors. It’s like, “Are you a husband or a father?”

TR: How do you decide what to pursue?

+  Paul McWIlliams says to read this item

The opinions expressed in the following interview are very well-aligned with some of the points I discussed in my “Micropower Revolution” report. The old strategy of throwing more transistors at design challenges simply does not work when power efficiency (MIPS per mW) is the design goal.

“Interview: Samueli Sees More Focus on Design Skill

By Linley Gwennap

The microprocessor industry is facing a huge challenge now that foundries can no longer double the transistor count every two years at the same cost. Where transistor budgets once increased almost for free, moving to the next node now adds cost. After 50 years of continual improvement in transistor count, the industry must adapt to a new environment. To help guide us through these changes, we turned to some very smart and successful technologists. The first interview in this series is with Henry Samueli, chairman and chief technical officer of Broadcom.

During the interview, Samueli commented, “It’s funny, for the last 20 years, our engineers are always complaining, ‘I don’t have time to optimize my current design because I’m too busy porting it to the next node.’ Now I’m saying, ‘Stay in the same node and just make your design better. Come up with a new architecture, new circuit designs, new algorithms, and optimize your design as opposed to optimizing your manufacturing.’ I think the slowing of Moore’s Law is going to benefit the design companies like Broadcom and hurt the manufacturing companies.”

Samueli has laid down the gauntlet to his engineering teams to focus on making their designs more efficient, not through simple transistor scaling or by throwing more transistors at the problem, but by using new design techniques. These techniques can include new algorithms, new instruction sets, new microarchitectures, and new circuit designs. Even without changing the process node, such techniques can improve efficiency by perhaps an order of magnitude (equivalent to three or four generations). Instead of racing to the next node, engineers will now race against their competitors to find the next clever improvement.”