DeepStream 7.1 Python RTSP In/Out Sample Crashes with cudaErrorIllegalAddress on Jetson Orin Nano

Accelerated Computing Intelligent Video Analytics DeepStream SDK
,
1

Hello NVIDIA team,

I am testing the DeepStream Python sample deepstream-rtsp-in-rtsp-out on a Jetson Orin Nano Developer Kit.

The pipeline starts correctly, runs for a while, but eventually crashes with the following CUDA/TensorRT errors:

Observations:

  • The issue happens after running for some time (not immediately at startup).

  • I modified the sample to use software x264 encoder since Jetson Orin Nano supports software encoding.

  • I replaced the default PGIE with a YOLOv8 TensorRT engine (custom model).

  • Errors consistently point to cudaErrorIllegalAddress (700) and failed TensorRT execution (enqueueV3).

System Information:

Device: Jetson Orin Nano Developer Kit

JetPack / L4T: JetPack 6.x

DeepStream SDK: 7.1.0

TensorRT: 10.3

CUDA: 12.6 (Driver + Runtime)

cuDNN: 9.0

nvcc:
nvcc: NVIDIA (R) Cuda compiler driver
Cuda compilation tools, release 12.6, V12.6.68
Build cuda_12.6.r12.6/compiler.34714021_0

error_log.txt (8.0 KB)

deepstream-app version 7.1.0
DeepStreamSDK 7.1.0
CUDA Driver Version: 12.6
CUDA Runtime Version: 12.6
TensorRT Version: 10.3
cuDNN Version: 9.0
libNVWarp360 Version: 2.0.1d3

9

please refer to this faq for the error “Failed in mem copy”.

10

Thank you for pointing me to the FAQ. I went through it but couldn’t find the exact solution for my issue. Since I’m using a software encoder, could you please guide me on resolving the “Failed in mem copy” error? It would also be very helpful if you could share a reference code example for using the software encoder along with NVMM memory.

11

As written in the faq, the fix is adding “copy-hw=2” for nvvideoconvert. Here is a sample cmd.

gst-launch-1.0 videotestsrc ! video/x-raw,format=RGBA ! nvvideoconvert copy-hw=2 ! video/x-raw\(memory:NVMM\),format=RGBA ! nveglglessink

please refer to this topic for SW encoding.