The same process goes on in the integrated-circuit business. The first case that comes to mind is militarized ICs. The Pentagon had its own rigid demands: high performance, resistance to extreme temperatures and radiation hardness. These demands drove a great deal of innovation in processes, devices and even processor architectures-anyone remember the RCA 1802 or the MIL-1750A CPU? No, I didn't think so.

But as the military faded as a customer, interest in all those developments waned. It turned out that these needs had become so specialized that they were no longer particularly useful outside their niche.

A similar process has taken place in the DRAM business. Once upon a time, DRAM was a technology driver. Intel was once a DRAM vendor that happened to make some logic chips. Not that many years ago the great Japanese fabs, with their superb process technology and their unmatched capital budgets, were dedicated to a DRAM-first strategy. Each new process node was introduced first as a DRAM process, and then a logic process was derived from it.

But with its need for tiny, high-valued capacitors creating very peculiar physical structures on the die, and with its indifference to complex interconnect stacks, DRAM rapidly diverged from the needs of logic designers. Today DRAM processes live in a different world from logic processes, overlapping only in occasional, expensive embedded-DRAM modules that are introduced late in a process' product life. And the DRAM business is, well . . .

A question worth asking as designers troop off to the Intel Developer Forum is whether the same thing is now taking place in the world of X86 microprocessors. For all practical purposes, there are only two vendors, neither happy with their earnings in the CPU business. The design of the chips themselves has increasingly diverged from the rest of design practice: domino logic, self-timed circuits, calculated confrontations with design rules and analysis tools that simply don't exist in the rest of the world.

Are the X86 processors also headed down the road toward technical marginalization, to become a specialty only practiced by a closed community? That might explain the diversity of subjects on tap for discussion at IDF.