Different ZED X framesrates across different Jetson machines

I have two jetson machines running systems that record data from multiple ZED X cameras.

On Machine 1, the framerate according to get_current_fps() is consistently 13-15 FPS, while on Machine 2 the framerate is consistently 9-11 FPS.

Machine 1 is using four cameras while Machine 2 has only three, so I’d expect the framerate to be higher on Machine 2.

The compute hardware is identical, both are:

Both systems are running identical code in a Docker image, and the ZED SDK version is the same, 4.1.3.

Could you give me some advice for other troubleshooting steps I could to uncover the cause of the discrepancy in framerates?

Any help is greatly appreciated.

Thanks!

Hi @andrew.stringfield
we are testing this behavior internally and will get back to you as soon as possible.

@andrew.stringfield can you confirm that you are using the same power mode (nvpmodel) on both devices?

@Myzhar yes, both machines are operating in MAXN power mode.

When I look at the journal logs while the recording application is running, I’m seeing the following Argus error a lot:

Aug 29 15:18:43 collection-2 nvargus-daemon[1168]: (Argus) Error AlreadyAllocated: Device 0 (of 1) is in use (in src/api/CameraProviderImpl.cpp, function createCaptureSessionInternal(), line 272)
  • On Machine 1, the error appears every now and then - every 30 seconds or so
  • On Machine 2, the error is logged constantly - several times per second

Could this be linked to the issue?

Thanks!

This can be a piece of useful information.
Can you confirm that the configuration is identical?
Same cable length, same GMSL2 ports, etc.

@Myzhar sure thing, config is as follows:

Machine 1

  • 3x ZED X connected to GMSL ports 2, 4 and 5 using 1.5-metre cables
  • 1x ZED X connected to GMSL port 7 using a 5-metre cable

Machine 2

  • 3x ZED X connected to GMSL ports 2, 4 and 5 using 1.5m cables

The configuration is the same for both machines, with Machine 1 having an additional camera connected via a 5-metre cable.

Hi @andrew.stringfield
we are trying to replicate your configuration to find a reason to the behavi or that you described.
Can you please run the command sudo ZED_Diagnostic --dmesg and send the report file to support@stereolabs.com?

@Myzhar I’ve generated a log and emailed to support, thanks. The diagnostic was run inside the Docker container.

I realised that both containers are actually running ZED SDK 4.0.8, not 4.1.3.

However, I’ve confirmed that the containers on both Machine 1 and Machine 2 are still running the same SDK version.

Thanks!