The International Technology Roadmap for Semiconductors (ITRS) just got out of
the fast lane.
The 2002 version, unveiled earlier this month, for the first time in several years
makes no change in the timetable for introducing new chip technology nodes. In
prior roadmaps, the introduction of leading edge nodes had been accelerated by a
year each time over earlier forecasts.
"Now the ITRS is back on schedule with the introduction of each new
semiconductor node coming on three year intervals up through 2016," summed up
Yuri Matisoo, director of technology programs for the Semiconductor Industry
Association, San Jose, Calif.
Almost from the time the chip roadmaps started in the last decade, each new node
was projected on a three-year pattern, except for the speedup the last two
The roadmaps for the next decade also bring microprocessors and DRAMs back into
sync, with both reaching the next semiconductor node at the same time. The last
fast-track versions had MPUs and ASICs a year ahead of DRAMs. Now the two chip
categories will be marching in technological lock-step.
For the record, here is the reaffirmed roadmap timetable for new process
technologies: 90nm for 2004, 65nm in 2007, 45nm in 2010, 32nm in 2013 and 22nm in 2016.
Actually, IBM Corp. and Intel Corp. will still be rushing the roadmap a bit by
introducing their first 90nm chips in the second half of 2003. But the
technology savants are probably right in forecasting that 90nm is a year away
for the bulk of the semiconductor market.
What does the chip technology crystal ball mean for systems builders and the
For one, it portends a slightly less hectic pace of bleeding-edge chips
overtaking the market. And with many electronic products facing restrained
sales, that might not be a bad thing. It should also lead to a little less of the
supply quandary plaguing vendors and customers that try to juggle inventories becoming
obsolete much sooner than anticipated.
Still an every-three-year jump into a new chip node is an impressive schedule.
It keeps pace with Moore's Law. "We were running ahead of Moore's Law for a
while. Getting back on track is a just a reality check," asserted SIA's Matisoo
A major factor in taking ITRS off steroids is that process technology is coming
much harder nowadays. The race from quarter micron all the way to 0.15 micron
(150nm) was accelerated because the same cornerstone lithography technology of
248-nm krypton fluoride laser exposure tools could handle each new generation.
Chipmakers didn't have to debug and qualify a new lithography system,
facilitating the transition to each new node.
Now several new lithography systems -- 193nm argon fluoride laser tools, upcoming
157nm tools, and the exotic EUV (extreme ultraviolet) -- are needed for the new
nodes, requiring a longer learning curve to ramp up. New materials and new chip
designs to cope with incredibly tiny line geometries complicate the shift to new
A side effect of the bringing ITRS back to reality is a more somber approach to
tackling the myriad of technical challenges to reach each new node.
days in some ways were the industry's technology bubble. Now we know the road
ahead doesn't come easy. There's no cruise control into the future.