IBM has first 7nm chip and leapfrogs over competitors
Chip makers are now producing 14nm processors, and the next big project for Intel and other chip makers has been the 10nm chip. IBM, in its announcement today, has upended the chip industry's development path.
A 7nm chip will hold about four times as many transistors in the same area as a 14nm chip, said Richard Doherty, research director of Envisioneering, a technology assessment and market research firm. In terms of chip development, IBM has "moved the field goal out," he said.
"For IBM to conquer 7nm without stopping at the 10nm that Intel is supposedly tackling, means that IBM has secured the future two steps out," Doherty said.
To create its 7mn test chips with functional transistors, IBM made "fundamental changes" in the materials used to create the chip as well as adopting new lithography techniques, said Mukesh Khare, vice president of semiconductor technology for IBM Research, in an interview.
Those technologies include the use of a silicon-germanium alloy, which improves the flow of electrons. IBM first described this technique in 2004, and said at the time it would lead to chip improvements in 10 years. A big advance in creating the 7nm chip was the use of extreme ultraviolet lithography. Optical lithography, which is now used in building chips, has a wavelength of 193nm, but extreme ultraviolet lithography (EUVL) has a wavelength of 13.5 nanometers, which carves much sharper patterns on silicon.
"IBM research is making sure that smaller transistors and scaling can continue," Khare said. "We want to make sure that high performance computing needs for IBM systems can be met for a very long time." IBM didn't disclose its timetable for commercial development, but said it is pushing the limits of technology "beyond" 7nm.
For 50 years, Moore's Law reliably predicted the doubling of transistors on a chip every two years. But as chip sizes reach the limits of physics, production costs have increased, and chip makers have been searching for new technologies to keep advancing computing performance.
A key issue will be whether IBM can bring its technology to commercial production, said Roger Kay, an analyst at Endpoint Technologies. For IBM, what follows "is the hard part" because of the cost and complexity of the equipment needed for mass production, he said.
IBM isn't going it alone. Last year, it announced a $3 billion, five-year investment in chip technologies and is working with Global Foundries and Samsung, among others. Its researchers are also working at the State University of New York's Poly's NanoTech Complex in Albany.
Intel hasn't detailed plans for 7nm chip but is expected to release a 10nm chip as later this year.
Richard Fichera, an analyst at Forrester, said IBM's effort doesn't guaranteethatb it will beat Intel to a commercial product. But if IBM continues to work with its partners and stay active in this area, it will have technologies to keep their Power chips and mainframe technology "very competitive."