Project Proposal

Summary

The Jetson Tegra K1 is a mobile computing development board which is an early prototype that promises to combine a Quad-core CPU setup which uses ARM Cortex-A15's with a kepler GPU which has 1 SMX and 192 cores. Both the CPU and Kepler GPU seek to deliver more computation per power than their predecessors. It is not always possible to characterize the power consumption of a system however the form factor provided by the Nvidia Jetson K1 allows just that. I seek to measure the power consumption of several parallel and multicore algorithms and compare them to their serial implementation in order to see if the benefits of parallelism are also present when adding power consumption into the equation.

The Challenge

The challenge is first chronicling how to work with a new development board produced by Nvidia. In addition I'm not a hundred percent sure how I'm going to go about measuring power in an accurate fashion. In addition I might work on improving the power efficiency of certain algorithms. What I hope to learn is what sorts of workloads present a challenge for mobile platforms. Providing adequate computing power while meeting the power requirements to run on a system which restricts the power and the cooling environment that these processors exist on can be daunting challenge.

Resources

The resources I will be using is a Tegra K1 board provided by NVIDIA it is the version with a 32-bit ARM Cortex-A15 QuadCore setup.Also there is the Tegra K1 White Paper and the Kepler architecture paper which I will review. In addition I will look over papers describing the necessity for a system to be power efficient.

Goals & Deliverables

PLAN TO ACHIEVE

Evaluate the power efficiency of the Tegra K1 board for several standard benchmarks. So have graphs as well as insight based on the tests that I perform. What will dictate how well the project goes is how accurately am I able to measure the power consumption and how relevant benchmarks used are. In addition since we have quite a bit of time I might have time to actually implement a power efficient version of some popular graphics algorithm.

HOPE TO ACHIEVE

Evaluate the power efficiency of the Tegra K1 board and also implement a power efficient algorithm using OPENGL ES.

Platform of Choice

Jetson Tegra K1 development board. I plan to be using C/C++ and OPENGL/OPENGL ES unless the board operates in a different fashion this is what I expect to be using. To really characterize power efficient we want to use as low level of a language as we can.