Hi, I’m the Toradex AI Assistant.
I might make mistakes. Our support team will step in if needed. If I answered your question, please tick the “Solution” checkbox. For follow-up questions to me, please use @ToradexAI in your post.
VPU Support Status in Torizon OS 6 and 7
You’re correct that there are currently limitations with hardware-accelerated GStreamer plugins in Torizon containers. Based on the available information, VPU support is not officially available in either Torizon OS 6 or Torizon OS 7 (1).
Current Status
The GStreamer packages inside Torizon containers don’t have support for utilizing the VPU on i.MX8 hardware, which explains why plugins like imxcompositor-g2d and v4l2h264enc are missing (1). This results in software-only video processing with high CPU usage.
Workaround for Torizon OS 6
There is a proof-of-concept container example available for VPU usage that was tested on Torizon OS 6.X (1). However, this example:
- Was intended as a proof-of-concept rather than officially tested
- Requires freezing the Debian package feed to a past snapshot for compatibility
- May not work correctly on Torizon OS 7.X
Hardware vs Software Processing
When using software-only processing, you’ll experience significantly higher CPU usage. For example, users report CPU consumption around 74-90% for video playback and up to 160% for streaming (1), compared to the much lower usage (typically under 15-20%) when hardware acceleration is working properly.
Alternative Approach
If hardware acceleration is critical for your product evaluation, you might consider using the Reference Multimedia Image with Yocto-based BSPs, where the VPU-accelerated GStreamer plugins are available (2). This approach provides access to hardware-accelerated elements like vpuenc_h264 and vpudec that you’ve successfully tested.
Future Support
Toradex has plans to re-package the GStreamer packages in containers for proper VPU hardware acceleration support, but there’s no specific timeline available for when this will be implemented (1).
For your product evaluation, you may need to weigh the benefits of containerized applications in Torizon against the current limitation of software-only video processing, or consider the Yocto-based approach for hardware-accelerated multimedia capabilities.
Best regards, Toradex AI Assistant.