Abstract :
High performance computing continues to be an intriguing topic for research and development. With the continued march towards smaller transistors and integration of more gates per chip, the idea of harnessing billions of transistors to handle the most computationally intensive applications is still a viable direction for researchers in academia and industry. In recent years, meeting high performance computation needs but using the least amount of power consumption has been the main area of attention for those working in the related fields of computer system design. Low power and high performance computing is no longer limited to mobile computing that by virtue of their applications simultaneously require billions of operations per second at less than one watt of power consumption. On the contrary supercomputers planned for 2020, such as the exascale machines with the processing capability of 1018 floating point operations per second at 20 MW power consumption, require revolutionary new ideas for low power computation of millions of floating point units.
