The lack of consensus over taxonomy and semantics contributes to the turmoil and rapid change that characterizes the IC industry today.

Indeed, what is this SOC thing? I have often asked, in various settings at the outset of a discussion, ýWhat exactly do you mean by system-on-a-chip?ý To give one answer, the currently accepted definition of a system-on-a-chip is a single IC with a processor, memory, and some other dedicated functions or special interfaces. Everyone talks about these complex ICs as if they're something new. However, system-level chips have been around for a long time. The first ones were never considered as system-level chips, but as microcontrollers with extended capabilitiesýseveral examples being the Motorola M68HC08 and the Intel 8051 series of microcomputers that came ready equipped with an 8-bit processor, a few kilobits of memory, and programmable peripherals. Although these early parts didn't sport the millions of gates that reside on many of today's chips, they certainly did meet all of the nominal requirements for classification as a system-on-a-chip.

All of this classifying leads us back to the crux of the problem: the industry generates such a huge quantity of words and phrases in attempting to meet the need to differentiate one circuit from the next, the descriptions often overlap upon each other. There are several obvious reasons for this morass of creativity; some people feel that what they are doing is so unique that everyone else's words simply don't do justice to their exquisite results. Only a carefully crafted set of words would be adequate and appropriate. In other cases, the available words just don't sound right. For instance, when some clever marketing person invented Application Specific Integrated Circuit, the resulting acronymýASICýwas a word you could pronounce. ASIC, however, quickly became synonymous with gate array and thus commanded even more words to further delineate (if not clarify) the entire issue.

Another place that is starting to confuse itself with its own nomenclature is the area of intellectual property. Now that we have silicon capacities exceeding 10 million gates and dropping into 0.18-micron geometries, we must redefine the notion of system-level integration once again. These very much larger integrated circuits won't be built as a single, flat IC, but as part of a collection of various pieces of previously designed functionsýor to use a term growing in popularity, intellectual property. Among the names suggested for these integrated functional units are: cores, system-level-macro components, macro functions, function blocks, macro cells, function cells, system-level-building-blocks, and many others as well. The Virtual Socket Integration Association (VSIA) calls these parts ývirtual componentsý or VCs and describes the design process as virtual prototyping. The plethora of names and descriptions for the pieces of intellectual property used to construct a highly complex IC, then, indicates the relative immaturity of the entire industry.

Finally, we don't even agree on the definition of the final product. One term for the resulting IC is system-level integration or SLI technology. Another term coming into common misuse is system-on-a-chip. Fooled you, didn't I? In reality, we still haven't gotten to an actual system-on-a-chip, because the ýsystemý includes such components as power control and user interface parts. Currently these components still require special packaging for the power devices and special interfaces for the user. For those systems that are connected to some other part of the world, the connection has be robustýimplying not static or voltage sensitive or tolerantýand of a fairly analog nature, characteristics not usually associated with a 0.18-micron processor operating on 2.5 volts.

One appealing aspect of the English language is its flexibility or adaptability. Over time, even in an industry like ours filled with jargon and three (or more)-letter acronyms, the language mostly helps us to communicate. As we move toward a set of system functions within a single piece of silicon, we will also see the eventual convergence toward a single set of descriptors for the internal components that constitute an total assemblage of functions on a single chip.