The World’s Best Performing AI Processor for Edge Devices

Home Landing Pages The World’s Best Performing AI Processor for Edge Devices

Discover more about Hailo AI Processors

Unprecedented AI compute power of 26 Tera-Operations Per Second (TOPS)

Best-in-class power efficiency with typical power consumption of 2.5W

Highest cost-efficiency (TOPS/$) compared with existing solutions

“This processor helps us overcome all the limitations”

Aviram Shmueli
Blue White Robotics , CTO & Co-Founder

Xavier NX results are using batch=8 (while Hailo-8 and Jetson Nano are using batch=1) and that Jetson Nano is limited to FP16 (while Hailo-8 and Xavier NX are INT8). Nvidia results for batch=1 and INT8, respectively, are expected to be lower.
Hailo-8 figures are based on SDK version 3.12.0 (November 2021), measured at room temperature on a single Hailo-8 device through PCIe interface on a Hailo evaluation board (system host: Intel® Core™ i5-9400 CPU @ 2.90GHz)
FPS & power figures for Nvidia Jetson Nano and Xavier NX are sourced from https://developer.nvidia.com/embedded/jetson-benchmarks and https://github.com/NVIDIA-AI-IOT/jetson_benchmarks/tree/master/benchmark_csv, retrieved 12/07/21

NN Model	Input Resolution	FPS	Power (W)	FPS/W
Classification
ResNet-50 v1	224×224	1,372	3.6	385
ViT_base	224×224	168	3.1	53
Object Detection
YOLOv5m	640×640	242	5.6	43
YOLOv8m	640×640	140	4.7	30
YOLOv11m	640×640	102	3.4	30
Semantic Segmentation
stdc1	1,024×1,920	58	3.0	19

Product Brief

Blog

When choosing an AI object detection network for edge d…