Performance Summary

The performance results of benchmarking the Isaac ROS packages on both Jetson and x86_64 platforms are below:

Node	Input Size	AGX Orin	Orin NX	Orin Nano Super 8GB	x86_64 w/ RTX 4090
AprilTag Node	720p	178 fps 6.3 ms @ 30Hz	116 fps 9.4 ms @ 30Hz	123 fps 8.6 ms @ 30Hz	596 fps 0.86 ms @ 30Hz
Freespace Segmentation Node	576p	3340 fps 1.7 ms @ 30Hz	2530 fps 1.5 ms @ 30Hz	2140 fps 1.9 ms @ 30Hz	3500 fps 0.44 ms @ 30Hz
Depth Segmentation Node	576p	41.4 fps 80 ms @ 30Hz	28.1 fps 98 ms @ 30Hz	–	105 fps 25 ms @ 30Hz
FoundationPose Pose Estimation Node	720p	1.54 fps 780 ms @ 30Hz	–	–	9.56 fps 110 ms @ 30Hz
DNN Stereo Disparity Node Full	576p	72.5 fps 17 ms @ 30Hz	42.1 fps 26 ms @ 30Hz	–	350 fps 2.1 ms @ 30Hz
DNN Stereo Disparity Node Light	288p	304 fps 5.9 ms @ 30Hz	143 fps 9.6 ms @ 30Hz	–	350 fps 1.6 ms @ 30Hz
Stereo Disparity Node	1080p	118 fps 11 ms @ 30Hz	78.1 fps 14 ms @ 30Hz	83.8 fps 13 ms @ 30Hz	943 fps 1.6 ms @ 30Hz
Rectify Node	1080p	800 fps 2.8 ms @ 30Hz	572 fps 3.3 ms @ 30Hz	595 fps 3.8 ms @ 30Hz	2500 fps 0.57 ms @ 30Hz
TensorRT Node DOPE	VGA	30.8 fps 37 ms @ 30Hz	15.5 fps 55 ms @ 30Hz	20.8 fps 51 ms @ 30Hz	298 fps 5.3 ms @ 30Hz
Triton Node DOPE	VGA	31.2 fps 340 ms @ 30Hz	15.5 fps 55 ms @ 30Hz	22.2 fps 490 ms @ 30Hz	277 fps 4.7 ms @ 30Hz
TensorRT Node PeopleSemSegNet	544p	489 fps 4.6 ms @ 30Hz	258 fps 7.1 ms @ 30Hz	269 fps 6.2 ms @ 30Hz	619 fps 2.2 ms @ 30Hz
Triton Node PeopleSemSegNet	544p	216 fps 5.5 ms @ 30Hz	143 fps 8.2 ms @ 30Hz	–	585 fps 2.5 ms @ 30Hz
DNN Image Encoder Node	VGA	339 fps 13 ms @ 30Hz	375 fps 12 ms @ 30Hz	–	480 fps 6.0 ms @ 30Hz
Occupancy Grid Localizer Node	~50 sq. m	19.6 fps 57 ms @ 30Hz	8.36 fps 130 ms @ 30Hz	9.02 fps 120 ms @ 30Hz	50.1 fps 8.5 ms @ 30Hz
H.264 Decoder Node	1080p	188 fps 7.3 ms @ 30Hz	–	–	596 fps 2.6 ms @ 30Hz
H.264 Encoder Node I-frame Support	1080p	402 fps 12 ms @ 30Hz	–	–	412 fps 3.2 ms @ 30Hz
H.264 Encoder Node P-frame Support	1080p	465 fps 11 ms @ 30Hz	–	–	596 fps 2.0 ms @ 30Hz
Nvblox Node	–	4.94 fps 34.0 ms	4.94 fps 155 ms	4.93 fps 87.4 ms	4.94 fps 200 ms

Graph	Input Size	AGX Orin	Orin NX	Orin Nano Super 8GB	x86_64 w/ RTX 4090
AprilTag Graph	720p	178 fps 9.1 ms @ 30Hz	111 fps 11 ms @ 30Hz	120 fps 11 ms @ 30Hz	596 fps 1.4 ms @ 30Hz
Freespace Segmentation Graph	576p	40.3 fps 79 ms @ 30Hz	27.6 fps 98 ms @ 30Hz	31.8 fps 55 ms @ 30Hz	102 fps 30 ms @ 30Hz
Centerpose Pose Estimation Graph	VGA	44.8 fps 43 ms @ 30Hz	29.0 fps 51 ms @ 30Hz	29.8 fps 50 ms @ 30Hz	50.2 fps 14 ms @ 30Hz
DOPE Pose Estimation Graph	VGA	27.3 fps 54 ms @ 30Hz	15.2 fps 73 ms @ 30Hz	–	186 fps 12 ms @ 30Hz
DNN Stereo Disparity Graph Full	576p	89.8 fps 19 ms @ 30Hz	35.2 fps 34 ms @ 30Hz	–	350 fps 5.8 ms @ 30Hz
DNN Stereo Disparity Graph Light	288p	184 fps 14 ms @ 30Hz	128 fps 14 ms @ 30Hz	–	350 fps 5.2 ms @ 30Hz
Stereo Disparity Graph	1080p	111 fps 15 ms @ 30Hz	72.2 fps 18 ms @ 30Hz	77.4 fps 18 ms @ 30Hz	692 fps 4.6 ms @ 30Hz
DetectNet Object Detection Graph	544p	55.4 fps 37 ms @ 30Hz	25.7 fps 45 ms @ 30Hz	33.0 fps 43 ms @ 30Hz	262 fps 11 ms @ 30Hz
RT-DETR Object Detection Graph SyntheticaDETR	720p	56.5 fps 29 ms @ 30Hz	33.3 fps 40 ms @ 30Hz	37.3 fps 37 ms @ 30Hz	450 fps 5.5 ms @ 30Hz
TensorRT Graph PeopleSemSegNet	544p	436 fps 10 ms @ 30Hz	212 fps 13 ms @ 30Hz	224 fps 13 ms @ 30Hz	587 fps 3.7 ms @ 30Hz
SAM Image Segmentation Graph Full SAM	720p	2.22 fps 390 ms @ 30Hz	–	–	14.6 fps 74 ms @ 30Hz
SAM Image Segmentation Graph Mobile SAM	720p	8.40 fps 120 ms @ 30Hz	2.22 fps 240 ms @ 30Hz	2.22 fps 230 ms @ 30Hz	62.5 fps 22 ms @ 30Hz

Live Graph	Input Size	Nova Carter
Data Recorder Live Graph 4 Hawk Cameras	1200p	22.4 fps (per stream avg) 0 dropped frames (avg)
Multicam Visual SLAM Live Graph 4 Hawk Cameras	1200p	30.1 fps
DNN Stereo Disparity Live Graph 3 Hawk Cameras 1x Full ESS and 2x Throttled Light ESS	1200p	Full: 30.2 fps Light: 15.2 fps (avg)
Perceptor Graph 3 Hawk Cameras	1200p	Nvblox ESDF: 9.45 fps Nvblox Mesh: 2.63 fps Visual Odometry: 30.0 fps

Methodology

The performance numbers provided above are measured using Isaac ROS Benchmark. Configuration files for reproducing results are provided in the launch scripts for each benchmark here.

Frames per second (fps) are measured using the maximum sustained framerate for the accelerated computational pipeline on each listed platforms. Performance measured includes input node → graph of node(s) under performance test → output node where the publishing rate of the input node is auto-tuned to discover the peak throughput dropping <5% of the frames. Average fps computed over 5 runs, discarding minimum and maximum runs; latency measured at 30hz publishing rate matching sensor capture rate.