Home / Opinions / 40 Years and Counting / 40 Years and Counting

40 Years and Counting

The Right Hand Turn - Power Efficiency

As Pat sees it, there have been three key stages in the development of microprocessor design. The first era was that of experimentation. At the start of it all, during the 60s and 70s new companies were springing up and collapsing all over the place. Everyone had new and exciting ideas about 'what are the materials, physics and chemistry {needed} to build these things'.

During the 80s, as methods became refined and a few big names started to establish themselves, the second phase of microprocessor design began. Now it was all about ramping up manufacturing capabilities, putting serious money into further research, and getting to the point where companies could 'deliver in high volume in a predictable way'. Around this time Pat made the prediction that Intel would reach 10GHz CPUs in twenty years time but, as he puts it now, ' I was probably wrong on that one by 20 years or so!'

The next era started to occur around the year 2000. With the world becoming ever more concerned about energy consumption, for all its negative impacts, the previous doctrine of 'more is always better' had to be revised to reflect the changing nature of our times. CPUs could no longer just get bigger and bigger, they needed to be more efficient for Moore's law to continue. This became known as the right hand turn.

However, if you look back at the actual products Intel were producing at the time this new focus on energy efficiency seems quite laughable. The Pentium 4 had just been introduced and it spectacularly failed to even come close to being what you might call power efficient. Of course, Pat is actually referring to the research stage of development so it would be a number of years before the fruits of this new way of thinking would come to be.

Power of Compatibility

Back in the early 90s, in his then position as lead architect, when he had just finished working on the 386 and was now concentrating his efforts on the 486, Pat recalls how he had been desperate for a load of the legacy logic that Intel had included in its chips from days gone by to finally be dropped. It took up valuable silicon and surely the world had moved on from ten years ago?

Unfortunately, or so it seemed at the time, Pat came up against marketing who insisted that it was needed. Looking back, though, Pat realises why he was in fact wrong and that maintaining backwards compatibility has always been a very important factor in Intel's success. It's all to do with something called the software spiral.

This is the phenomenon whereby software gets developed to support a certain set of hardware creating a little bubble of commerce. If a new generation of hardware is developed to be compatible with this software it already has an established market to slot into so has a guaranteed level of commercialism. Continue this cycle a few times and you get to the point where the amount of software using a certain architecture grows exponentially. This is known as the software spiral and ignoring the commercial power of it would have been disaster for Intel.

As Pat puts it 'we had billions of dollars of software ready to run for our next generation processors even before we announced them if we could deliver on that promise of compatibility. It was all about the architecture, the investments of the industry and the software'.

In fact, now here's a statistic for you, a perfect demonstration of the software spiral, or at least an abstraction thereof, is the 386. Intel didn't stop producing this CPU until September 2007, some twenty years after it first started production. With clients like NASA still demanding these basic and reliable CPUs for mission critical components of its space craft, Intel had to maintain production until long after it would seem the chip had had its day. The same goes for all that software out there. Intel simply can't afford to abandon it.

comments powered by Disqus