DeepStream 7.1 CUDA Synchronization Issue on WSL2 with RTX 4060

Accelerated Computing Intelligent Video Analytics DeepStream SDK
, , ,
1

DeepStream 7.1 CUDA Synchronization Issue on WSL2 with RTX 4060

Hi everyone,

I’m running into a persistent issue with my DeepStream 7.1 pipeline on WSL2 (Ubuntu 22.04) using an NVIDIA RTX 4060 GPU. Despite experimenting with various memory type configurations for pipeline elements like nvv4l2decoder, nvvideoconvert, and nvstreammux, I consistently encounter a CUDA synchronization error. I’d really appreciate any insights or suggestions from the community!

Issue Summary

  • Problem: The pipeline fails due to a CUDA memory synchronization error.
  • Error Message:
ERROR: nvdsinfer_context_impl.cpp:1954 Failed to synchronize on cuda copy-complete-event, cuda err_no:999, err_str:cudaErrorUnknown
  • Environment:
    • DeepStream Version: 7.1
    • OS: WSL2 (Ubuntu 22.04 on Windows 11)
    • GPU: NVIDIA RTX 4060
    • NVIDIA Driver: 576.40
    • CUDA Version: 12.9
    • TensorRT Version: 10.9.0.34

Pipeline Overview

I’m working with a standard DeepStream pipeline for processing RTSP video streams:
rtspsrc → rtph264depay → h264parse → nvv4l2decoder → nvstreammux → nvinfer → nvvideoconvert → nvdsosd → fakesink

Memory Type Configurations Tried

I’ve tested several nvbuf-memory-type settings for key elements:

  1. Configuration 1:
  • nvv4l2decoder: PINNED
  • nvvideoconvert: CUDA_PINNED
  • nvstreammux: CUDA_PINNED
  • Result: Pipeline starts but crashes after some time with the CUDA error.
  1. Configuration 2:
  • nvv4l2decoder: UNIFIED
  • nvvideoconvert: CUDA_DEVICE
  • nvstreammux: CUDA_DEVICE
  • Result: Pipeline fails to start, even with a single camera.
  1. Configuration 3:
  • nvv4l2decoder: DEVICE
  • nvvideoconvert: CUDA_DEVICE
  • nvstreammux: CUDA_DEVICE
  • Result: Pipeline starts but eventually fails with the same CUDA error.

Additional Observations

  • The error occurs more quickly when multiple camera streams are added to the pipeline.
  • I’ve confirmed the NVIDIA driver (576.40) is up to date, but I’ve come across mentions of potential compatibility issues with newer drivers.
  • Increasing WSL2’s memory allocation didn’t resolve the issue.

Questions

  1. Has anyone successfully run DeepStream 7.1 on WSL2 with similar hardware? If so, what memory type configurations worked for you?
  2. Are there known compatibility issues between NVIDIA driver version 576.40 and DeepStream 7.1 on WSL2?
  3. Could this be tied to WSL2’s virtualized GPU access? Are there specific workarounds for this environment?

Any advice or experiences you can share would be greatly appreciated. Thanks in advance!

3

Please refer to the following document for installation. However, we do not recommend using DeepStream on WSL2. Deepstream is still in alpha version on WSL and has many issues. It is recommended to use native Ubuntu 2204 to install DeepStream

1 Like