October 19, 2006
Carnegie Mellon University has been awarded a six-year, $4.2 million grant from the Defense Advanced Research Projects Agency (DARPA) to create a new type of reconfigurable integrated circuit for chip manufacturers.
"Through our new Center for Memory Intensive Self-Configuring Integrated Circuits (MISC IC), we will create new nanoscale chip designs and architectures that will be intelligent enough to fix themselves," said Ed Schlesinger, center director and head of Carnegie Mellon's Electrical and Computer Engineering (ECE) Department.
The race to produce higher-performance integrated circuits is limited today by the cost of production and the challenges faced by the traditional paradigm of making components smaller and smaller. Carnegie Mellon researchers will work to create technology that seeks to break this bottleneck by integrating mechanical probes with integrated circuits in a design that allows for the reconfiguration of integrated circuits. In addition, these systems will allow for the integration of memory, storage and processing technologies with minimal additional cost to competitive global chipmakers.
Schlesinger said that so far most chipmakers have relied mostly on processes that shrink the critical dimensions on a chip--and the transistor size--thereby speeding up the processing of data. But shrinking critical dimension size is becoming more and more difficult.
In the 1990s, shrinking, known as scaling in the industry, led to faster speeds. "Today, though, circuits are packed so closely together and feature sizes are so small that various effects are limiting performance," said Jim Bain, associate director of the Data Storage Systems Center (DSSC) and a professor of ECE. The new center will also benefit from the technical expertise and financial support of the DSSC in ECE.
Carnegie Mellon researchers believe that there are new approaches that need not rely on the shrinking of transistor size alone to continue the improvement in integrated circuit performance. The new DARPA- and industry-funded research center is developing new materials and circuit configurations that combine the strengths of integrated circuits, microelectromechanical systems and information storage to create reconfigurable systems.
"This physical reorganization of the chip itself will drive many changes in both hardware and software development," Schlesinger said.
Today, only integrated circuits that can be produced and sold by the millions are generally economical to produce because of high production costs. Integrated circuits are then customized, usually by the software running on them. With circuits that can reconfigure themselves, there exists the possibility of economically producing complex systems as well as improving the performance of application-specific integrated circuits at the low cost now so endemic to the production of more general-purpose integrated circuits, Carnegie Mellon researchers said.
"In the future it is possible to imagine producing economical integrated circuits for applications that demand relatively low volumes of integrated circuits," said Schlesinger. "Our new collaborative center will help create the underlying technology that could help change the integrated circuit paradigm and sustain momentum in the $227 billion global chip sector for years."
Source: Chriss Swaney, Carnegie Mellon News