SAN FRANCISCO— Henry Samueli spun my head around so fastI thought it would snap off. He explainedwhy the worst possible thing that couldhappen to the electronics industry might bea good thing for Broadcom, the company heco-founded.
Moore's Law is slowing down, and its endappears to be in sight, he said. Nosurprise. Back in May he became thefirst semiconductor executive I ever heardfrankly acknowledge this increasing apparentreality. Then he added a new head-snappingtwist:
- From our perspective [thissituation] gives us more breathing roomto be clever about design. Most people runto the current process node as fast asthey can. That's going to change. Insteadof running to the next node, you will comeup with new architectures and circuitdesigns, and that will create moreopportunities on the design side, bringingmore value to a company like Broadcom.
Laughing, I asked, “So Moore's Law isending… and that's a good thing forBroadcom?”
We both laughed. “Yes, in a way it is,” hesaid.
While Samueli argues Broadcom could expandits slice, the overall pie will slow itspace of growth, a prospect that quicklysobered up our conversation.
- There will be a slowdown in rate ofinnovation in products [end users]buy. The smaller/cheaper/faster [dynamic]is definitely going to slow andpotentially 15 years from now even stop –then it's a matter of leveraging thesystem as a whole rather than the endproduct.
Samueli is not alone in speaking franklyabout this trend. In a keynote at theInternational Electron Devices Meeting (IEDM), one of the topgatherings on the future of semiconductors,Geoffrey Yeap, vice president of technologyat Qualcomm, painted a similar picture.
Chipmakers face “more material/process costand design complexity… to meet productspecifications for low leakage/power andhigher performance,” he wrote in a paper forthe IEDM proceedings.
“This positive feedback loop drasticallyaccelerates the increase in die cost($/mm2)… making area scaling lessattractive,” Yeap wrote. “We are gettingdangerously close to this inflection pointas the scaling box with four sides of speed,density, power, and cost is becoming smalleras we march towards the 7 nm node.”
Engineers are researching several areas inhopes of countering or at least slowing downthe trend. They include new backendmaterials and processes, multiple 3D chipstacking efforts, and industry collaborationon extreme ultraviolet lithography, 450mmwafers, and design optimizations of allsorts, Yeap wrote.
To read more, go to “Long life for the 28 nanometer node.”