Moore's Law drives Intel R&D
As a company, Intel aims to live up to the expectation of co-founder Gordon Moore that it should produce new process technology every two years.
This expectation has since become known as Moore's Law, after an observation made in 1965 that the number of transistors on an integrated circuit doubled every 24 months.
Addressing the Intel Developers Forum, in Beijing, China this week, Mark Bohr, a senior fellow in Intel's Technology and Manufacturing Group, said this means a continued focus on research and development.
So, while Intel will later this year begin production of a new generation of processing chips using 45nm technology, researchers are already studying a generation based on 22nm and developers are readying a generation based around 32nm, he said.
The latter will be released in 2009 and the former most likely in 2011. This should be down to 16nm in the next 10 years and 8nm in the next 16 years, Bohr noted.
Each chip generation has advantages over the previous in areas other than just smaller lithography. Bohr explained that Penryn and Silverthorne, the 45nm generation, use a Hafnium-based Hi-k Metal Gate transistor in the place of silicon oxide. They also have twice the transistor density of their 65nm predecessors and provide 20% faster switching speeds. Switching power is also down 30% and oxide leakage power by a factor of 10.
The next generation may feature tri-gate transistors, Indium Antimonide quantum well transistors and carbon nanotube interconnects. Tri-gate transistors offer improved performance and offer a lower source drain than linear transistors, Bohr said, adding that carbon nanotubes have "very interesting conducting capabilities".
However, there is a hurdle to cross, he noted. Current production technology will be adequate to manufacture 32nm chips, but not the generations beyond that.
As a result, Intel is studying extreme ultraviolet technology to take it beyond 32nm and keep the promise contained in Moore's Law, Bohr concluded.