Predicting power scalability in a reconfigurable platform

Beckett, P 2007, Predicting power scalability in a reconfigurable platform, Doctor of Philosophy (PhD), Electrical and Computer Engineering, RMIT University.


Document type: Thesis
Collection: Theses

Attached Files
Name Description MIMEType Size
Beckett.pdf Thesis application/pdf 3.35MB
Title Predicting power scalability in a reconfigurable platform
Author(s) Beckett, P
Year 2007
Abstract This thesis focuses on the evolution of digital hardware systems. A reconfigurable platform is proposed and analysed based on thin-body, fully-depleted silicon-on-insulator Schottky-barrier transistors with metal gates and silicide source/drain (TBFDSBSOI). These offer the potential for simplified processing that will allow them to reach ultimate nanoscale gate dimensions.

Technology CAD was used to show that the threshold voltage in TBFDSBSOI devices will be controllable by gate potentials that scale down with the channel dimensions while remaining within appropriate gate reliability limits. SPICE simulations determined that the magnitude of the threshold shift predicted by TCAD software would be sufficient to control the logic configuration of a simple, regular array of these TBFDSBSOI transistors as well as to constrain its overall subthreshold power growth. Using these devices, a reconfigurable platform is proposed based on a regular 6-input, 6-output NOR LUT block in which the logic and configuration functions of the array are mapped onto separate gates of the double-gate device.

A new analytic model of the relationship between power (P), area (A) and performance (T) has been developed based on a simple VLSI complexity metric of the form ATσ = constant. As σ defines the performance “return” gained as a result of an increase in area, it also represents a bound on the architectural options available in power-scalable digital systems. This analytic model was used to determine that simple computing functions mapped to the reconfigurable platform will exhibit continuous power-area-performance scaling behavior.

A number of simple arithmetic circuits were mapped to the array and their delay and subthreshold leakage analysed over a representative range of supply and threshold voltages, thus determining a worse-case range for the device/circuit-level parameters of the model. Finally, an architectural simulation was built in VHDL-AMS. The frequency scaling described by σ, combined with the device/circuit-level parameters predicts the overall power and performance scaling of parallel architectures mapped to the array.
Degree Doctor of Philosophy (PhD)
Institution RMIT University
School, Department or Centre Electrical and Computer Engineering
Keyword(s) Transistors
Versions
Version Filter Type
Access Statistics: 319 Abstract Views, 608 File Downloads  -  Detailed Statistics
Created: Tue, 30 Nov 2010, 14:10:48 EST by Keely Chapman
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us