PALO ALTO, Calif. – Big hurdles in silicon scaling and battery technology stand in the way of huge opportunities in mobile systems, said an ARM executive in a keynote here.
"Silicon scaling will end at some point, and I think it's coming sooner than many people think," said Simon Segars, general manager of ARM's physical IP division in a keynote at the annual Hot Chips event here. What's more, "we really need a new battery technology," he said.
With silicon atoms measuring a fraction of a nanometer in diameter and process technology approaching single nm digits, "you can only scale so far before we need other materials like III-V semiconductors," he said.
Difficulty delivering production-quality extreme ultraviolet (EUV) lithography is already creating problems. Chip makers must use complex double-patterning techniques with today's immersion systems while they wait for EUV that may be required for the 14nm node, Segars said.
"You need to produce 200-300 wafers an hour, and today's EUV machines can do about five wafers per hour now," said Segars. "Some people question whether it ever will be mainstream--lots of R&D still needs to go into it," he added.
Design issues loom large, too. 4G modems could be 500-times the complexity of 2G versions, requiring dedicated data-processing engines.
The need for more performance and power is driving up the complexity in multiple power domains and timing closure, he said. Nevertheless, the ARM exec promised advances including by 2015 Cortex A15 processors fully coherent across multiple CPUs and GPUs.
Battery technology looks equally challenging, increasing only about 11 percent a year, far behind the pace of Moore's Law. Even maintaining that sluggish rate "will require some exotic materials such as silicon alloys or carbon nanotubes—batteries are really rubbish," he said.
On the other side of the hurdles are huge mobile opportunities. "All the numbers are big," said Segars, noting sales of 280 million smartphones last year and a market of four billion cellular subscribers.
"Although we have made fantastic progress there are a few issues ahead and the future won't be like the past," he warned.
Delays in EUV threaten progress toward 14nm, Segars said.