Operand-Value-Based Optimizations To Reduce Processor Power Consumption

Web Published:

Researchers at Princeton University led by Professor Margaret Martonosi have developed a novel technology to dramatically reduce power consumption in commercial microprocessors. Princeton is currently seeking industrial collaborators to commercialize this technology.


State of the art processor designs have been pushed towards 64-bit word-widths because of the large address space needs in current applications. Although full 64-bit addresses and operations are indeed sometimes needed, arithmetic operations on much smaller quantities are still more common. Across the SPECint95 benchmarks, over half the integer operation executions require 16 bits or less. Based on this date, Professor Martonosi and her group have developed a hardware mechanism that dynamic ally recognizes these narrow-width operations.


This optimization reduces power consumption by using operand-value-based clock gating to turn off portions of arithmetic units that will be unused by narrow width operations and results in a 50% reduction in the integer unit's power consumption for the SPECint95 and MediaBench benchmark suites. It is anticipated that additional work will lead to further power reductions within other units of the processor.


Applications would include a wide range of commercial microprocessors including high end systems to reduce the costs of chip packaging and heat dissipation and to extend the battery life in notebooks and other embedded systems. Moreover, increased demand for this type of technology is expected as the usage of larger word-width microprocessors becomes more prevalent in printers, routers, multimedia, etc.


Patent protection is pending.


For more information please contact:

             William H. Gowen
             Office of Patents and Licensing
             Princeton University
             4 New South Building
             Princeton, NJ 08544-0036
             (609) 258-6762
             (609) 258-1159 fax

Patent Information:
Computers and Software
For Information, Contact:
John Ritter
Princeton University
Margaret Martonosi
David Brooks