• Hardware Platform (Jetson / GPU)
Jetson AGX Thor (ARM64) and x86+dGPU (both reproduce on DeepStream 8.0)
• DeepStream Version
DeepStream 8.0 (nvcr.io/nvidia/deepstream:8.0-triton-multiarch)
Comparison baseline: DeepStream 7.1 (nvcr.io/nvidia/deepstream:7.1-triton-multiarch) where issue is not reproduced.
• JetPack Version (valid for Jetson only)
38.2.2-20250925153837
• TensorRT Version
DeepStream container bundled TensorRT version (please see container startup banner).
x86+dGPU: 10.15.1.29-1+cuda13.1
Jetson: 10.13.3.9-1+cuda13.0
• NVIDIA GPU Driver Version (valid for GPU only)
x86+dGPU: Driver Version: 590.48.01 (NVIDIA GeForce RTX 4060)
Jetson: 580.00
• Issue Type (questions, new requirements, bugs)
Bug
• How to reproduce the issue?
We prepared a minimal standalone reproducer (nvmm_appsrc_enc_test.cpp + Makefile) with no project/business logic.
Flow:
appsink -> external NVMM pool (gst_nvds_buffer_pool_new) -> appsrc -> nvv4l2h26xenc -> parse -> qtmux -> filesink
Important behavior:
- DS 8.0:
--frames-limit 32: no failure--frames-limit 33: reproducible failure
- DS 7.1:
- no failure with
--frames-limit 33
- no failure with
- Reproduced with both:
nvv4l2h264encnvv4l2h265enc
Run command (DS 8.0, from nvmm_appsrc_enc_test/ directory):
docker run --rm -it --gpus all -e NVIDIA_DRIVER_CAPABILITIES=compute,video,utility \
-v "$PWD":/work -w /work \
--entrypoint /bin/bash \
nvcr.io/nvidia/deepstream:8.0-triton-multiarch \
-lc 'make clean all && ./nvmm_appsrc_enc_test --frames-limit 33'
Run command (DS 7.1, comparison):
docker run --rm -it --gpus all -e NVIDIA_DRIVER_CAPABILITIES=compute,video,utility \
-v "$PWD":/work -w /work \
--entrypoint /bin/bash \
nvcr.io/nvidia/deepstream:7.1-triton-multiarch \
-lc 'make clean all && ./nvmm_appsrc_enc_test --frames-limit 33'
Typical failure log on DS 8.0:
ERROR from rec-enc: Failed to process frame.
DEBUG: /dvs/git/dirty/git-master_linux/3rdparty/gst/gst-v4l2/gst-v4l2/gstv4l2videoenc.c(1930): gst_v4l2_video_enc_handle_frame (): /GstPipeline:record-pipeline/nvv4l2h264enc:rec-enc:
Maybe be due to not enough memory or failing driver
We can attach full source and complete GST debug logs if needed.
nvmm_appsrc_enc_test.zip (6.1 KB)