Joseph and Thread - Semiconductor Business News Article - "Move To Speed Up [Process Technology] Roadmap Triggers Industry Fight"
Note Intel vs. rest of Industry. (Guess who's side I'm on? ;-))
supersite.net
Move to speed up roadmap triggers industry fight
By Jack Robertson and J. Robert Lineback
MUNICH - A fight has broken out in the semiconductor industry over what, if any, changes should be made in its technology roadmaps. The problem is simply that many people now believe the roadmap no longer does a good enough job in helping companies to plan their production and development schedules. Semiconductor managers complain that production tools haven't been ready when they're needed, partly because leading-edge companies are moving faster now with their device shrinks than the roadmap specified. This makes it harder to gauge when key production tools, materials, and process are needed to launch the next-generation ICs. So these managers want to pick up the pace on the roadmaps and set formal targets for new process generations on a two-year cycle rather than the current three-year pace.
But other industry managers worry about what they see as an accelerating cost of speeding up the technology timetable if two-year cycles are used in the next revision of the roadmap, which is officially called the 1999 International Technology Roadmap for Semiconductors. The opponents of a two-year cycle believe that not only will it require more investments by struggling companies but also will feed on itself and result in a premature end to some fundamental device technologies and production techniques.
This all will be debated in mid-April when chip technologists from all over the world gather here in Munich prior to the Semicon Europa trade show to work on the 1999 international schedule and argue over whether to speed up the three-year cycles used in U.S. roadmaps since 1994. The new global roadmap, being drafted by technology working groups coordinated by the U.S. Semiconductor Industry Association (SIA), is due out by the end of the year.
"This is a highly controversial issue and it's driven by many different objectives and positions at individual companies," says a concerned Ronald Dornseif, the analyst who tracks semiconductor manufacturing technology at Dataquest in San Jose. "For the leading companies - in particular, Intel - it makes a lot of sense to accelerate the pace because it will thin out the competition," he says. "But for the industry as a whole, it makes no sense at all."
The Dataquest analyst, along with a number of industry managers, worries that a two-year roadmap cycle will further accelerate the pace of device shrinks as chip companies attempt to beat the industry's roadmap targets. "This is a self-fulfilling prophecy, and it's equivalent to putting the industry on a treadmill where it will eventually suffer a heart attack," Dornseif warns.
National's Padmanabhan
By going to a two-year cycle, the industry will get less profit from each new IC generation because the life of the processes and production equipment will be cut short, Dornseif maintains. "The lifecycle will not be sufficient to get an acceptable return on the investments," he complains. The accelerated pace of device shrinks, Dornseif claims, also will speed up the time when the silicon transistor runs out of steam - possibly around the 0.03- or 0.02-micron generation. The danger here, he says, is that "accelerating the roadmap will speed up the end of the transistor long before a replacement will be ready."
But the advocates of change say the inability of the current roadmap to accurately predict developments already has been costly to the industry. They point out that the 1997 update and an unscheduled 1998 revision to the SIA roadmap resulted in a full year being sliced out of the schedule because chip makers were able to shrink devices at a much faster rate than originally expected.
The 1998 revision started out as a minor adjustment to the 1997 SIA roadmap, but by the time it was ratified three months ago, it had shifted process nodes ahead by one year in recognition that developments had outrun projections in just 12 months (see story in the Jan. 15 publication).
"The pace has been pretty aggressive in the past few years, and the next five years will be crucial," says Alex Oscilowski, chief operating officer of Sematech's international subsidiary in Austin, Tex. "There are concerns about the investments [needed to maintain this accelerated pace] given the economics faced by both the equipment suppliers and semiconductor companies in the past couple of years," he says. To keep the industry on track, he figures, more cooperation will be needed between companies and international consortia.
Concerns over growing costs are uppermost in the minds of most chip makers still trying to recover from three years of anemic growth and losses stemming from an oversupply of ICs. In Japan, for example, many of the semiconductor companies are against moving the roadmap to a two-year cycle, says Hiroyoshi Komiya, executive vice president of that nation's Selete (Semiconductor Leading Edge Technology) consortium.
"The roadmap should keep to its schedule of three-year intervals for future generations," Komiya argues.
Roadmap officials at SIA's San Jose headquarters believe the chip industry is split on the proposal to shift to two-year cycles in the 1999 roadmap. Those for and against the accelerated pace are "evenly divided among companies" in both Japan and the United States, says Juri Matisoo, SIA's vice president of technology programs who oversees the roadmap activities.
But some people question whether changing the length of the cycle between major technology nodes would have that much effect on the rate of technology change. "I would predict we would continue at about the same pace we've been, and I don't think anyone has yet repealed Moore's law," contends Bob Payne, senior vice president for strategic technologies at VSLI Technology Inc. in San Jose.
Others are actively pushing for a two-year cycle in the upcoming international roadmap. "The question is whether two years is aggressive enough," notes Peter Rickert, technology platform manager at Texas Instruments Inc. in Dallas. "I think the answer for the past two-to-three years is probably no. For the next two-to-four years, things are going to get more difficult in terms of transitioning from technology [to] having the equipment ready," he predicts. "But stretching it out to three years is definitely out of the question. TI is supporting the two-year cycle."
A number of major chip houses are backing the two-year cycle partly because they believe wafer fabrication tools are not keeping up with their own requirements. Lithography tops the list of concerns, according to TI's Rickert and others. Some technology managers believe the mismatch between roadmap targets and the accelerated pace of device shrinks resulted in 193-nm lithography tools not being ready for pilot lines when they were needed in late 1998 and early this year.
"The 193-nm [exposure tools] are just now becoming available and we are waiting for the second-generation systems that will be needed for production," complains Gobi R. Padmanabhan, senior vice president of technology research and development at National Semiconductor Corp. in Santa Clara, Calif. In addition to lithography, the National vice president is worried about design automation, metal interconnect processes, chip testers, and flip-chip packaging technologies. They all lag the needs of device makers, he adds.
Padmanabhan is another backer of the two-year cycle. He believes it should be used until the middle of the next decade. Beyond that, he adds, a three-year cycle would be better because there is a "need for major developments" to push beyond the 0.10-micron feature size.
At the Munich meetings in mid-April, roadmap working groups will be hit by a number of proposals, according to industry sources, including an aggressive and controversial attempt to set two-year intervals between process generations throughout the next decade. If the 1999 roadmap adopts that pace, leading DRAM makers will be aiming to launch new memory chips with 0.10-micron half-pitch feature sizes in 2003. That's two years earlier than the 2005 now set in the 1998 revision of the roadmap.
The proposed two-year cycle would also pull up the 0.07-micron DRAM generation to 2005 instead of 2007. The aggressive proposal aims to put 0.05-micron DRAM technology into production in 2007 - five years ahead of the 2012 time frame set in the 1997 U.S. roadmap.
The shorter cycle also would speed up the technology targets in microprocessor technology. Isolated gate lengths in MPUs would shrink to 0.10 micron in 2001 vs. 2002 in the 1998 revision of the roadmap. The 0.07- and 0.05-micron MPU targets would be pushed up to 2003 and 2005, if the two-year march is maintained.
Accelerating the roadmap would raise problems, say other industry executives. Sematech's Oscilowski and others worry about the pace of industry growth early in the next decade if the current pace is not maintained. "It would certainly be a challenge," he comments. "But the biggest challenge we face is how we make the best use of limited resources.
"We want to move forward as fast as we can, but [what we do] also needs to be affordable," Oscilowski added. "I think you will see a little more effort going into the affordability side of the roadmap this year than you might have seen in the past." |
