15-418 Spring 2014

Tegra K1 Power/Energy Analysis

| Delvis Taveras | dtaveras@github.com | delvistaveras@gmail.com


Week Goals
Dates start Monday and end Sunday
Mar 31 to Apr 6 Choose project
Apr 7 to 13
- Read/Review papers justifying the trends towards low power computing
- Read/Review papers on latest trends of mobile computing
- Find benchmarks and testing harnesses used for GPUs

Apr 14 to 20
- Go to techshop and see if they have current sensors if not order one
- Setup current sensor and have it communicate with ardunio or raspberry pi
- Have raspberry pi/ardunio sample the current as fast it can and communicate the data to the computer to generate the time vs power plot
- Code GUI using QT which reads in serial data and generates a real time "time vs power" plot

Apr 21 to 27
- Go through and document the basic power consumption information of the system.
- Found a need to collect and compare data and so we must add an option on the current plot to allow one to do this.
- Need to add the ability to save to the testing harness.

Apr 28 to May 4
- Another sanity check can we improve our tests are there any especially power hungry algorithms
- Can we find odd relationships between power and speed does the fastest algorithm always consume the least power.
- Test the power consumption of compute bound applications, of memory bound applications as well as combination of cpu and gpu stressing algorithms.

May 5 to May 6
- Update the website one last time and prepare final presentation
- What anomalies did we find?
- Were there any limits as to how accurately we could measure the power
- Any suggestions for those wanting to measure the power consumption of algorithms correctly?
- Is the Tegra K1 a game changer?

News & Updates May 7 2014

Nvidia Tegra K1 and second current sensor and Arduino held on until the end.