Billions and Billions of transistors
07/01/1997
Billions and billions of transistors
In the Gordon Moore interview, conducted by West Coast Editor Ed Korczynski in this issue (see "Moore`s Law extended: The return of cleverness," Industry Insights, page 364), Moore speculated on "...what do you do with a billion transistors of circuitry?" This is an interesting question, with many ramifications, and a challenge for application designers in all market segments.
Historically, hardware had to keep up with software. Software demands drove the MPU development for faster, more functional chips. Software developers built databases and communications logic that made chip designers reach for higher and higher megahertz. Sub-100-MHz chips are now museum pieces, with 600-MHz models available. Graphics software and CAD/CAE applications became mainstays of many industries with full three-dimensional modeling that required the highest-performance workstations.
The communications arena was the same, where functionality required gigabit/sec bandwidth to be easily affordable with super-fast switching. Heap on the communications overload caused by the Internet, which is just in its infancy, and the chip has its work cut out for it.
However, a change looms just over the horizon. Although the demand for faster processors is still driving chips to smaller feature sizes, the fact is that the chip-masters will be throwing the gauntlet back to the software designers by giving them nearly unlimited raw circuit power (in the form of a billion transistors on a single chip). This kind of virtually unlimited logic palette will spawn a wide variety of new ideas and concepts and will change the basics of the industry. Bright new designers from fabless companies could proliferate entirely new chips for (currently) unpredictable applications. The system-on-a-chip could heat up the quest for a computer on a chip for less than $20 by 2002.
All these forces will make the marketing of chips more difficult to predict. A billion-transistor-chip will give rise to applications that are not yet formulated, and dropping price points will enable them to penetrate markets where they are not now found. More consumer products will emerge, and the time from introduction to commodity status will shrink.
Also, as the logic possibilities become unlimited, more complex and "system-like" designs will ensue, and the issue of intellectual property rights will enter in a major way. Presaging this is the recent Digital suit against Intel regarding the Pentium processors.
How the industry will survive all this turmoil will probably be a result of something else Moore mentioned - "cleverness." One should never underestimate the cleverness factor; it is somehow magically built into our market-driven system.
As Moore points out, cleverness gets us around barriers, be they economic or technological. While it is true that companies must improve efficiency, they must also "grow the pie larger" by the return of cleverness, and by providing customers with innovative short time-to-market products.
We are approaching the era where many of the low-hanging apples have been picked. Now it`s time for market-ingenuity and new levels of cleverness to profilerate, both in chip design and manufacturing. It is a challenge the technical community worldwide always seems to meet, and this time will surely be no exception.