Intel Corp. has all but completed its shift to a 0.25-micron production process, and by this time next year the company will begin producing even smaller, cheaper, faster, and more capable chips using a 0.18-micron process.
Almost all of Intel's processor production has moved to a 0.25-micron process, said Paul Otellini, Intel's executive vice president and general manager of the Intel Architecture business group. The older 0.35-micron technology is still being used only on some Pentium MMX (P5) CPUs and on some of the earliest Pentium II (P6) processors, he said.
Intel has ended "wafer starts" on the Pentium Pro processor and will sell off warehouse and in-process inventories as applications shift to the Pentium II Xeon, Otellini said.
"The production on the high end has all moved over to the Pentium II Xeon," he said.
With smaller geometries allowing Intel to put more chips on a given wafer, production cost per chip falls. The smaller devices are faster, yet consume less power. And at a given chip size, more circuitry -- such as Level 2 cache memory -- can be built on each chip, improving performance.
Intel's transition from 0.35 to 0.25 microns was less painful than the company expected, allowing Intel to speed up the introduction of faster CPUs this year and next. Intel is on track with this product roadmap and plans to offer an IA-32 microprocessor in 1999 and a 64-bit microprocessor (Merced) in 2000, each using the 0.18-micron process.
Some industry observers have surmised that the complex Merced processor demands the lower power consumption of a 0.18-micron process.
The transition to 0.18-micron manufacturing is ahead of schedule, according to Otellini, so production of the first parts to use the process is being pulled in at least one quarter, to the first half of 1999.
One result of the technology shift is a reduction in this year's capital spending, Otellini said. As 0.18-micron production starts, 0.25-micron production slows. As a result, Intel will not have to purchase additional capital equipment until next year.
Another result is a slowdown in the transition from 200-millimeter (8-inch) to 300-mm (12-inch) wafers. The ability to put more chips on a wafer reduces the need to use larger, more expensive wafers.
"There seems to be a little bit of a setback in 300-mm," said Janet Ramkissoon, an executive at Quadra Capital, in New York.
The shift to 300 mm is "a little farther out than before," Otellini agreed.
Intel Corp., in Santa Clara, California, can be reached at http://www.intel.com/.