# Comparing Energy Efficiency of CPU, GPU and FPGA Implementations for Vision Kernels

arXiv: https://arxiv.org/abs/1906.11879 · HTML: https://mlsanity.com/paper/1906.11879/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 31 May 2019.

## Built on by (most cited first, 8 of 8)

- [Treehouse: A Case For Carbon-Aware Datacenter Software](https://mlsanity.com/paper/2201.02120/) (2022)
- [HyDe: The First Open-Source, Python-Based, GPU-Accelerated Hyperspectral Denoising Package](https://mlsanity.com/paper/2204.06979/) (2022)
- [HAPM -- Hardware Aware Pruning Method for CNN hardware accelerators in resource constrained devices](https://mlsanity.com/paper/2408.14055/) (2024)
- [ImageHD: Energy-Efficient On-Device Continual Learning of Visual Representations via Hyperdimensional Computing](https://mlsanity.com/paper/2604.21280/) (2026)
- [THOR: A Generic Energy Estimation Approach for On-Device Training](https://mlsanity.com/paper/2501.16397/) (2025)
- [Accelerated Feature Detectors for Visual SLAM: A Comparative Study of FPGA vs GPU](https://mlsanity.com/paper/2510.13546/) (2025)
- [FPGA or GPU? Analyzing comparative research for application-specific guidance](https://mlsanity.com/paper/2511.06565/) (2025)
- [FPGA-Based Hardware Architecture for Contrast Maximization in Event-Based Vision](https://mlsanity.com/paper/2605.09581/) (2026)

## Abstract (arXiv)

Developing high performance embedded vision applications requires balancing run-time performance with energy constraints. Given the mix of hardware accelerators that exist for embedded computer vision (e.g. multi-core CPUs, GPUs, and FPGAs), and their associated vendor optimized vision libraries, it becomes a challenge for developers to navigate this fragmented solution space. To aid with determining which embedded platform is most suitable for their application, we conduct a comprehensive benchmark of the run-time performance and energy efficiency of a wide range of vision kernels. We discuss rationales for why a given underlying hardware architecture innately performs well or poorly based on the characteristics of a range of vision kernel categories. Specifically, our study is performed for three commonly used HW accelerators for embedded vision applications: ARM57 CPU, Jetson TX2 GPU and ZCU102 FPGA, using their vendor optimized vision libraries: OpenCV, VisionWorks and xfOpenCV. Our results show that the GPU achieves an energy/frame reduction ratio of 1.1-3.2x compared to the others for simple kernels. While for more complicated kernels and complete vision pipelines, the FPGA outperforms the others with energy/frame reduction ratios of 1.2-22.3x. It is also observed that the FPGA performs increasingly better as a vision application's pipeline complexity grows.