CMOS 'to stay' even when Moore's Law stops
CMOS scaling could end by 2020 but there may be no serious competitor to it from other nanotechnologies before 2030, according to Dennis Buss, chief scientist of major chipmaker Texas Instruments.
“There is no nanotechnology that will replace CMOS before 2030 and even then I am fairly sceptical,” said Buss in a keynote address at the SAME conference in Sophia Antipolis 50 years on from Jack Kilby’s demonstration of a integrated circuit. “The device that will replace CMOS has not yet been invented.”
Buss said he estimates the final generation of CMOS technology will have gates that are just 10nm across – the current leading-edge process has gates that can be as small as a little under 40nm. Buss noted that scaling had yielded density improvement of 130 fold since the commercial introduction of integrated circuits a little over 40 years ago. “There is only 4x left to go,” he warned.
By that time, a single chip could economically carry 100 billion transistors. “Even when CMOS struggles, think of what you will be able to do with hundreds of billions of transistors. So, it will be very hard to replace CMOS. However, if you consider some of the things you could put on top of CMOS, I think that is where nanotechnology will help us,” said Buss, pointing to novel memory technologies as possibly being the primary candidates for post-CMOS work.
Buss said the cost of process development has forced even the biggest chipmakers into alliances and design cost forces many companies off of leading-edge processes.
“The end of the brick wall is not so much a brick wall but more of a swamp. In a swamp, it just gets harder and harder to put one foot in front of the other. But, when we get to the end of Moore’s Law scaling, does that mean the end of the industry? The answer is clearly no. If you look at the application areas that will dominate electronics they don’t need a hundred billion or even a billion transistors,” Buss explained. “Medical electronics, energy generation and storage, power scavenging. Many of these applications are the ones we see as being the future of our industry.”
As a result, many companies are stepping back from trying to be among the first onto a process. TI itself is focusing more on mixed-signal designs that do not require minimum-size transistors. “Being a CMOS technology leader is becoming less of an advantage than it was ten years ago. It became a religion. But that view is changing.”