Investigation and development of energy saving techniques for modern x86 platforms: with a special emphasis on embedded environments

In recent times there is a growing emphasis on energy saving solutions for processor-based systems of all types, and there is a particular emphasis on energy-saving solutions to sustain battery life in portable and remote equipment. Some 75 percent of the world’s data processing solutions employ the x86 processor technology, in one form or another, as the underlying computing architecture. However, the ubiquitous x86 technology is characterized by high dissipation loss in relation to its computing power. Despite this significant drawback, the x86 technology is used in many applications, including embedded processor solutions. This research project is concerned with the development of a novel technical solution which will help to make x86 based embedded systems more energy efficient. The work focuses on the widely-used PICMG COM Express Standard. The proposed solution defines an intelligent interface, referred to as the “Embedded Controller Management Interface”, which is used to optimise the energy management for the embedded system. This “Embedded Controller Management Interface” is implemented by using a dedicated microcontroller to handle the various energy management tasks. The prototype implementation consists of an adapter which includes a FPGA and an ARM-7 microcontroller to perform the various measurements. The collected data can be accessed remotely via Ethernet and is provided to the DUT (device under test) using a standardised Application Programming Interface (API) or a standard bus system, such as the I2C. The work in this project includes background theoretical research for energy use in embedded systems, measurements of energy use in embedded systems, and the requirements specification and design for the special management interface board. The circuit implementations for some selected core modules are presented and test results for the research work to date are presented. The results of this research work show it is possible to monitor the electrical power consumption of an embedded module in order to perform instantaneous on-line energy profiling.