SDAA320 June   2026 AM625 , AM625SIP , AM62D-Q1 , AM62L , AM62P

 

  1.   1
  2.   Abstract
  3.   Trademarks
  4. 1 Introduction
    1. 1.1 Key Highlights:
    2. 1.2 Basic Concepts:
      1. 1.2.1 PipeWire Server
      2. 1.2.2 PipeWire Clients
      3. 1.2.3 Session Manager
      4. 1.2.4 Nodes, Ports, and Links
    3. 1.3 PipeWire Main Components
  5. 2 Linux Audio Stack
  6. 3 Build SDK Image with PipeWire Support via Yocto
    1. 3.1 Steps to Run Yocto Builds on Host
      1. 3.1.1 Prerequisites (One-time setup)
    2. 3.2 Clone the oe-layer Setup
    3. 3.3 Download and Apply PipeWire Patches
    4. 3.4 Build PipeWire Image
  7. 4 Setup PipeWire on Sitara Devices
    1. 4.1 Hardware
      1. 4.1.1 SK-AM62B-P1
      2. 4.1.2 TMDS62LEVM
      3. 4.1.3 AUDIO-AM62D-EVM
    2. 4.2 Configure EVM Boot Mode
      1. 4.2.1 SK-AM62B-P1
      2. 4.2.2 TMDS62LEVM
      3. 4.2.3 AUDIO-AM62D-EVM
    3. 4.3 UART Console Setup
    4. 4.4 Flash the SD Card Image
    5. 4.5 Booting EVM with SD Card
  8. 5 Use PipeWire
    1. 5.1 Check Service Status
    2. 5.2 Enable PipeWire and Wireplumber
    3. 5.3 Start PipeWire and WirePlumber
    4. 5.4 General PipeWire commands
      1. 5.4.1 List All Objects Currently in PipeWire Server
      2. 5.4.2 List Only Nodes
      3. 5.4.3 Inspect Specific Object
    5. 5.5 Play and Record Stereo Audio
  9. 6 Configuration
    1. 6.1 Sink and Source Configuration
    2. 6.2 WirePlumber Configuration
  10. 7 Performance Benchmarks
    1. 7.1 Latency
    2. 7.2 CPU and Memory Usage
    3. 7.3 CPU and Memory Usage with Resampling
    4. 7.4 Observations
  11. 8 Summary
  12. 9 References
  13. 10IMPORTANT NOTICE AND DISCLAIMER

Observations

The benchmark results are largely platform agnostic because the tested workload (5 concurrent audio streams) is relatively lightweight and does not significantly stress CPU, multicore scaling and DDR bandwidth . As a result, differences in CPU frequency, core count and DDR speed have minimal impact on CPU usage, latency, and memory occupancy. The small differences observed are likely within normal run-to-run variation due to factors such as scheduler timing, background activities, cache, and so on.