November 6, 2020 — Strategy and Tactics: Why People Get the End of Moore's Law So Wrong. Maxim: CEOs must make on-the-spot decisions on the basis of trust.
WildPhotons: Worrying won't stop the bad stuff from happening it just ...
G. Dan Hutcheson
Why People Get the End of Moore's Law So Wrong:For the umpteenth time, the future of Moore's Law has come into question by senior industry executives using projections from pundits. These show a historical decline in cost-per-gate (a gate is typically made up of 6 transistors, which is the metric I prefer because it's how Gordon laid things out originally) with a projected rise in the future. Now you probably remember that the same things have been said before at previous nodes. Still, in seeing such new projections, I know they were wrong before and I believe the new projections are on their way to being wrong again. Let me show you why... The first reason for erring is they take a static view rather than a dynamic, time domain, view. The second reason is the 'What you see is all there is' bias. They see a bottom at the last node, with a rise in the new node entering production, and assume that's the final cost point for that node.
What they do next is project that rise out into the future and conclude Moore's Law is dead. Now I should point out that Gordon himself made this very error in his first 'Moore's Law' paper in 1965. Four decades ago, I wrote that this U-turn in costs was the economic effect of being at the edge of the best technology at the time. Cross this edge and costs go up. I also said what engineers do is push this edge out in time. Let's look as some dynamic data that shows how this has played out over time.
Maxim: CEOs must make on-the-spot decisions on the basis of trust.
Three of Japan's best companies in the semiconductor industry are Advantest, JSR, and TEL. I once asked Nobu Koshiba, then head of JSR, why they were so successful when so many others failed. Without blinking an eye he said the key to their success, where other Japanese companies failed so badly, was the ability to make on-the-spot decisions with real commitments to follow-thru in front of the customer. Among customers, Nobu was famous for this in committing to develop ArFi and EUV resists, where the time-to-money horizon was unknowable. The best story I ever heard is the commitment Terry Higash made at the signing between TEL and IBM for Albany:
"Don't worry. We are partners now. I'll sign" —Terry Higashi
ChipChirpsTM — VLSI's Semiconductor Analytics report and more. Semiconductor sales continued to set seasonal records. But the IC weather cooled significantly, with OSAT, IoT, Auto, and Analog & Power dropping a notch. Focusing on the Logic picture this week... ChipChirps are short notifications from VLSI's analysts to alert clients about breaking news and events. They are exclusively available on VLSI's App for Android and iOS.
VLSIinsiders' cloudside chat videos:New: 4 Nov 4 and 28 Oct: Order Visibility, Red Flags, & the economy.
The Chip Insider® button in VLSI's App for Android and iOS to download or Click here for the website.
"When everything seems to be going against you ...
remember that the airplane takes off against the wind" —Henry Ford