HONOLULU With memory bits set to take up the vast majority of tomorrow's system-on-chip solutions, new forms of memory technology will be needed to keep power consumption under control, according to Yoshio Nishi, director of the Center for Integrated Systems at Stanford University.
Speaking here at the IEEE 2006 Silicon Nanoelectronics Workshop on Sunday (June 11) Nishi said: "By 2016, chips will be almost all memory, and almost all of the power consumption will be from memory."
For these reasons Nishi said that some currently well-favored memory technologies, such as phase change memory, could run up against power limitations. Nishi said he has doubts about the prospects for phase change memory (PCM) because it requires heat to reset the state of a memory bit. Phase change techniques may consume too much power as the bit density increases, he suggested.
Nishi's group at Stanford University is studying a form of resistance memory, called conductance-bridge, resistance-change memory, which includes metallic links between bits. A similar approach could be used for new forms of logic, he said.
A number of commercial companies are working on the resistive RAM concept including a startup, Unity Semiconductor Corp. (Sunnyvale, Calif.). Unity, founded in 2002, doesn't expect to come to market before 2008 or 2009.
In ten or 15 years, 3D circuits, which often connect logic, memory, and optoelectronics functions with vertical interconnections, will be an established technology, Nishi told about 200 participants at the two-day workshop, which precedes the larger Symposium on VLSI Technology which starts Tuesday.
The use of 3D technologies will "have a major impact on design and layout. We in the technology community need to start a dialog with the design community about 3D circuits and new forms of nanotechnology. We may find out something from the designers that we otherwise might not think about," he said.
New forms of non-volatile memory are often seen as replacements for flash, which faces scaling limitations. However, Nishi said volatile memory types, such as SRAM and DRAM, face their own scaling challenges. As it becomes more difficult to limit current dissipation in the off state, the memory work horses of the computer industry static RAM and dynamic RAM may be replaced by "new forms of universal memories," he said.
Before joining Stanford, Nishi worked at Texas Instruments, Hewlett-Packard, and Toshiba, managing silicon technology development.