SAN JOSE, Calif. -- The microprocessor industry is in a consolidation phase and should rally around extending the x86 architecture to new market spaces, especially for network and handheld systems, said a keynoter at the Microprocessor Forum here Wednesday (Oct 15).
"The future for micro architecture development is very bright," said Fred Weber, chief technology officer of Advanced Micro Device's computer CPU group.
"We have an incredible amount of work to do, but it's not in designing new instruction set architectures. ISAs are one of the least powerful and most costly levers we can turn," he added.
Weber, who disclosed AMD's plans for 64-bit extensions to the x86 at the same conference several years ago, declined to specify any new extensions AMD would like to see. However, he did suggest the need for work making the x86 ready for networking and handhelds.
"We are already seeing the x86 get significant use in storage area networking. In the future will it extend to the network itself and to ubiquitous handheld devices? I think so," Weber said.
Cisco Systems has requested both AMD and Intel extend the x86 to support big endian code used in its Internetworking Operating Systems. However in an interview with EE Times, Weber was noncommittal on such extensions. He noted the volumes of routers and switches are far below those of PCs, and chip makers must consider relative opportunity costs in focusing on one set of extensions over another.
As for handhelds, AMD recently acquired the Geode processor from National Semiconductor aimed at low-cost and mobile x86 systems. Although ARM-based architectures now hold sway in PDAs and cellphones, Weber said that could change.
"It's not clear to me that ARM has won the battle for the applications processor on the cellphone. You could say ARM and the Motorola Dragonball have won the PDA, but I don't buy it," he said.
Weber predicted that if cellphones become general-purpose computers, designers will leverage the installed base of PC software and hardware to drive them. However, that could require breakthroughs in display or voice-recognition technology, he said.
Weber reiterated his belief that putting multiple cores on a die will become a more important trend than multithreading individual processors. Nevertheless he said he expects individual processors will eventually go to two- and possibly four-threaded engines.