Well, last week, I was going to write my editorial about chips and “the end of Moore’s Law”, but I started and it turned out to be something else! Now it seems doubly appropriate, following the death this week of Andy Grove of Intel, who was such a big part of driving the Moore’s Law forecast to become true over many years. Grove was one of the founders of Intel and a big influence on its direction and strategy.
I loved a quote I saw the other day from Peter Lee – VP at Microsoft Research – who said, “there’s a law about Moore’s Law – the number of people predicting the end of Moore’s Law doubles every two years”. The battle simply for higher clock speeds has been more or less stopped for several years now and although chips get more complex, many aspects are constrained. As a result of the tiny features on the chip, at least in terms of silicon as the main substrate, it does look as though Moore’s Law will struggle to hold good.
What will the effect of that be? Well, it seems to me unlikely that at the end of Moore’s Law, which counts transistors, there will be a sudden stop to progress and acceleration in how things run. Constraints in simply scaling up silicon will be overcome by being more clever in how the silicon gets used. I’m not a chip specialist, but I’m intrigued by the idea that companies are working on how to re-programme chips on the fly using code to modify the hardware on FPGAs in real time, to allow different hardware architectures to be used depending on the demands of the system. Microsoft has a research project called “Catapult” (http://tinyurl.com/guodh9x) that uses this concept in server applications, particularly built around its Bing search engines. It will be intriguing to see if the technology makes it into other areas and client devices.
In the end, the technology development over my working life has been quite astonishing. In 1971, I was a trainee in an electronics development lab and saw my first chip. I was astonished to see a chip that had a complete Schmitt trigger (a circuit for generating a square wave from an analogue signal) inside. Up to then, a number of discrete components were needed. Of course, as I was young and obsessed with music, I immediately tried to use one to make a ‘fuzz box’ for my guitar, although it turned out that this wasn’t a very good way to do it!
In looking up some components for this article, I came across the TI 7414, which includes half a dozen Schmitt triggers and has a data sheet that was created in 1983, updated in 2002. I wonder how many other chips were designed that long ago and are still available in the market?
Returning to Andy Grove, he was responsible for a Comdex keynote that famously talked about the “Battle for Eyeballs” (i.e. that consumer’s spending follows where their eyeballs are focused – if they watch TV, they spend money on the sets and content – if they play games on PCs, they spend on those). That has been a useful concept for analysts, ever since.
Bob
