Abstract

Artificial intelligence (AI) has been shifted to the embedded devices known as edge devices. Component-level power is very important for the design and optimization of applications on edge devices to estimate energy consumption. Thus, accurate power measurements are needed for battery-powered systems. However, it is not straightforward. Because the behavior of a GPU is rather complex and not well documented. In this work, we report challenges getting power measurements using the built-in power sensor for an NVIDIA Jetson GPU device. We provide a method for true power and energy measurements of the kernels running on NVIDIA Jetson family GPUs.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call