I’m encountering a significant performance issue when using the ROS2 RTX Lidar Helper node. Enabling the Publish Full Scan option causes the PointCloud2 topic publish rate to drop to an unusable ~0.4 Hz
Environment
Isaac Sim Version: 5.0.0
Operating System: Ubuntu 22.04
GPU: NVIDIA GeForce RTX 4060 Ti (16GB VRAM)
NVIDIA Driver Version: 570.133.07
ROS 2 Distribution: Humble
Detailed Problem Description
The goal is to publish a complete scan from an RTX Lidar as a single PointCloud2 message. The ROS2 RTX Lidar Helper node has a boolean parameter, Publish Full Scan, for this exact purpose.
Expected Behavior: When Publish Full Scan is enabled, the node should accumulate data until a full rotation is complete and then publish the aggregated point cloud.
Observed Behavior: The /point_cloud topic is published at an extremely low frequency of about 0.4 Hz (one message every 2.5 seconds), making it unsuitable for most robotics applications.
Steps to Reproduce
The issue can be reproduced with a minimal setup:
Launch Isaac Sim.
Create a new RTX Lidar: Create > Sensors > RTX Lidar > [Any Model].
Use the ROS 2 graph generation wizard: Tools > Robotics > ROS2 OmniGraphs > RTX Lidar.
In the wizard, set Lidar Prim to the Lidar created in step 2.
Enable the Point Cloud checkbox and click OK to generate the OmniGraph.
In the Stage or Graph Editor, select the ROS2_RTX_Lidar_Helper node (located under /World/Graphs/ROS_LidarRTX_Graph by default).
In the Property panel for the node, check the Publish Full Scan checkbox.
Start the simulation by clicking the Play button.
In a ROS 2-sourced terminal, monitor the topic frequency:
ros2 topic hz /point_cloud
Observe the reported rate is approximately 0.4 Hz.
Troubleshooting & Analysis
I have performed several tests to isolate the bottleneck:
Internal Bottleneck Confirmation: The slowdown appears to be internal to Isaac Sim, not a ROS 2 communication issue. Enabling Show Debug View on the ROS2 RTX Lidar Helper node shows the debug point cloud rendering inside the viewport at the same slow rate (~0.4 Hz). This strongly suggests the bottleneck is within the helper node’s data accumulation/processing logic, before the data is published to ROS 2.
RMW Implementation: I changed the ROS 2 middleware from the default rmw_fastrtps_cpp to rmw_cyclonedds_cpp. This had no impact on the publish rate, further ruling out a network or middleware-specific problem.
Lidar Comparison: I replicated the setup using the older LiDAR (the lidar that is defined as a camera prim from Isaac Sim “4.2.0”). The exact same slow publishing behavior occurs. This indicates the issue may not be specific to the new RTX Lidar sensor but lies within the common ROS2 RTX Lidar Helper node logic that handles the “full scan” feature.
This “Slow PointCloud2 Publishing” issue in isaac sim 5.0.0 is a critical blocker preventing me from migrating.
For your convenience, I have attached a minimal USD file that demonstrates this issue immediately upon loading and playing the simulation. (The file is a result of the steps defined above. You may use the file or create a new stage and folllow the steps.)
Hi @BaSiSa. I tried the steps you provided and also the USD file you provided in Isaac Sim 5.0. I am not able to reproduce the issue. The publish rate of the point cloud is consistently 60Hz on my end.
Just for your reference, here is my output of nvidia-smi:
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 560.35.03 Driver Version: 560.35.03 CUDA Version: 12.6 |
|-----------------------------------------+------------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA RTX 5000 Ada Gene... Off | 00000000:01:00.0 On | Off |
| 40% 69C P2 154W / 250W | 4867MiB / 32760MiB | 52% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| 0 N/A N/A 2649 G /usr/lib/xorg/Xorg 288MiB |
| 0 N/A N/A 2809 G /usr/bin/gnome-shell 252MiB |
| 0 N/A N/A 3991 G ...seed-version=20250804-050057.034000 223MiB |
| 0 N/A N/A 2342782 C+G ..._build/linux-x86_64/release/kit/kit 3996MiB |
+-----------------------------------------------------------------------------------------+
Here is the output of nvcc --version:
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2025 NVIDIA Corporation
Built on Wed_Jan_15_19:20:09_PST_2025
Cuda compilation tools, release 12.8, V12.8.61
Build cuda_12.8.r12.8/compiler.35404655_0
Thanks for the reply.
On my first PC (RTX 4060 Ti), I installed nvcc 12.8 after noticing it wasn’t present initially. I also restarted the system after installation, but the issue still persisted.
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 570.133.07 Driver Version: 570.133.07 CUDA Version: 12.8 |
|-----------------------------------------+------------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA GeForce RTX 4060 Ti Off | 00000000:01:00.0 On | N/A |
| 0% 60C P0 37W / 165W | 3743MiB / 16380MiB | 32% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2025 NVIDIA Corporation
Built on Wed_Jan_15_19:20:09_PST_2025
Cuda compilation tools, release 12.8, V12.8.61
Build cuda_12.8.r12.8/compiler.35404655_0
I tried the same thing on a second PC (RTX 4090) which has nvcc 11.8 installed, and I observed the exact same issue there.
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 550.144.03 Driver Version: 550.144.03 CUDA Version: 12.4 |
|-----------------------------------------+------------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA GeForce RTX 4090 Off | 00000000:01:00.0 On | Off |
| 30% 48C P2 174W / 450W | 7732MiB / 24564MiB | 56% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2022 NVIDIA Corporation
Built on Wed_Sep_21_10:33:58_PDT_2022
Cuda compilation tools, release 11.8, V11.8.89
Build cuda_11.8.r11.8/compiler.31833905_0
I can share any information about my environment if that would help diagnose the problem.
Could anyone else spare about 5 minutes to try reproducing this issue? I would be very thankful.
I’m trying to determine whether the error is specific to my setups—e.g., a particular version of something or something I’ve installed—
@BaSiSa I am able to replicate this issue with the download link from the Isaac Sim documentation even without enabling “publish full scan”.
I tried an earlier commit and it doesn’t have this issue e1b0516f2e95da19be2d722ac7257c329e426237
But still I don’t recommend you to use pre-release commit as there are other issues…
Let me reach out to the internal team about this issue so we can fix it properly.
Thank you for confirming the issue and for reaching out to the internal team about it—I really appreciate your support.
I wasn’t able to build the repo at that specific commit. I also tried commit 2c6f5aefdb6322aa44951e71fcfb973ad413c774, but in that version the USD file I shared in my original post wasn’t working properly—the LiDAR wasn’t functioning.
I’m really looking forward to the next release with this fix—I can’t wait to try it out!
Thanks again for your time and help!
There is a performance issue on Isaac Sim so it cannot reach high publish rate and it will be addressed in Isaac Sim 5.1.
As a workaround, you can try setting skipDroppingInvalidPoints to True (if it is False by default). Then you should see an immediate 2-3x improvement in the full scan publish rate.
