PowerSensor3: A Fast and Accurate Open Source Power Measurement Tool

TL;DR

PowerSensor3 presents an open hardware/software toolkit for fast, high-resolution power measurement of PCIe devices and SoCs, achieving up to $20\,\mathrm{kHz}$ sampling with modular sensor boards and differential isolation. It combines a baseboard with dedicated sensor modules, an STM32F411 firmware stack, and a cross-platform host library (C++ with Python bindings) to deliver precise, real-time energy data and long-term stability with minimal calibration. The work demonstrates superior time resolution over vendor tools, enabling detailed GPU transient analyses, integration with Kernel Tuner for energy-aware auto-tuning, and effective SSD/power profiling; it also shows practical coverage across GPUs (RTX 4000 Ada, W7700), NVIDIA Jetson Orin, and Samsung 980 PRO SSDs, all at a cost under €100. These contributions offer a reusable, cost-efficient pathway to quantify and optimize energy efficiency in modern AI, HPC, and data-center workloads, potentially reducing the environmental footprint of large-scale computing. The combination of open hardware, open software, and validated case studies positions PowerSensor3 as a versatile tool for researchers and developers to adopt energy as a first-class metric in performance optimization.

Abstract

Power consumption is a major concern in data centers and HPC applications, with GPUs typically accounting for more than half of system power usage. While accurate power measurement tools are crucial for optimizing the energy efficiency of (GPU) applications, both built-in power sensors as well as state-of-the-art power meters often lack the accuracy and temporal granularity needed, or are impractical to use. Released as open hardware, firmware, and software, PowerSensor3 provides a cost-effective solution for evaluating energy efficiency, enabling advancements in sustainable computing. The toolkit consists of a baseboard with a variety of sensor modules accompanied by host libraries with C++ and Python bindings. PowerSensor3 enables real-time power measurements of SoC boards and PCIe cards, including GPUs, FPGAs, NICs, SSDs, and domain-specific AI and ML accelerators. Additionally, it provides significant improvements over previous tools, such as a robust and modular design, current sensors resistant to external interference, simplified calibration, and a sampling rate up to 20 kHz, which is essential to identify GPU behavior at high temporal granularity. This work describes the toolkit design, evaluates its performance characteristics, and shows several use cases (GPUs, NVIDIA Jetson AGX Orin, and SSD), demonstrating PowerSensor3's potential to significantly enhance energy efficiency in modern computing environments.

Figures (16)

• Figure 1: Schematic of a PowerSensor3 measurement setup.
• Figure 2: 3D rendering of PowerSensor3 with PCIe 8-pin, 20 A, 10 A, USB-C sensor modules, "Black Pill" module and display.
• Figure 3: Measurement setup for accuracy assessment.
• Figure 4: Power error for four types of sensors with dotted lines indicating min and max values per measurement point.
• Figure 5: Step response of PowerSensor3: load stepped from 3.3 A to 8 A plotted in ms scale (left) and $\mu$s scale (right).