eFuse is a patented technology that combines software algorithms and microscopic electrical fuses that will produce chips that can adapt on the fly to increase performance or avoid problems, according to IBM officials.
The technology can sense when the chip needs to increase performance or avoid a potential problem, and then can reconfigure the chip to meet the demands by tripping electrical fuses integrated into the chip. It can monitor and manage power consumption, repair problems and sense changes in demands on the chip.
Chip makers can use the morphing technology to alter chips for systems makers depending on the needs of end-users, officials said.
Bernard Myerson, IBM fellow, vice president and chief technologist of IBMs systems and technology group, touched upon the technology in March when Big Blue announced it was opening up its Power Architecture.
IBM Foundry customers can start using the eFuse technology now, in such jobs as testing and validating chips as they come off the wafer, spokesman Chris Andrews said. That work is currently done manually, using lasers to repair or remove imperfections.
With eFuse technology, the processor itself can do the testing and validation tasks, and make changes as needed, all of which reduces costs and the time to market for products, Andrews said. Future uses could include autonomic processors that can self-monitor, self-heal and reconfigure themselves dynamically after theyve been put into systems, he said.
eFuse can be used in a variety of chips, including IBMs Power5 and low-power SiGe (silicon germanium) processors for wired and wireless communication devices, Andrews said. IBM officials expect the technology to appear in chips for everything from cell phones to computers.
The technology is part of IBMs overall autonomic computing initiative, which is designed to remove as much human intervention as possible from the computing process.
"This strategy [currently] is in the hardware systems and software level," Andrews said. "Were adopting it at the chip level, bringing autonomic characteristics down to the chip level."