[Support]: Frigate + RADEON RX560 #17659
Replies: 3 comments 7 replies
-
This is not likely to be a Frigate bug if it works correctly with your nvidia setup. Without any logs it's impossible to help at this point. |
Beta Was this translation helpful? Give feedback.
-
Yeah, you'll want to look and see what is using system memory. I have multiple systems running with AMD GPU on latest and it has no issue |
Beta Was this translation helpful? Give feedback.
-
I reinstalled the dedicated RX560 GPU. Until now, I had been using the integrated Radeon GPU from the processor without major issues. However, after switching to the RX560, the system became completely unusable within seconds of starting Frigate. I didn't have time to capture any running processes or logs because the system simply froze completely. I'm currently in the process of restarting it to attempt another test. |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Checklist
Describe the problem you are having
Summary:
When using Frigate (latest stable version) with an AMD Radeon RX560 GPU on Ubuntu 22.04 Server, the ffmpeg process gradually consumes all available system memory, eventually leading to a complete system freeze. This issue does not occur when using an NVIDIA GPU under the same setup.
System Configuration:
OS: Ubuntu 22.04.4 LTS (Server)
Frigate Version: Latest stable (as of report date)
GPU: AMD Radeon RX560
GPU Driver: Default amdgpu driver available from Ubuntu’s official repositories
Cameras: 12 IP cameras (generic Chinese brand)
System RAM: 24 GB
FFmpeg Versions Tested: 5.0 and 7.0 — same memory leak behavior observed in both
Docker: [If applicable, insert method of deployment]
Details:
The issue starts shortly after launching Frigate with the RX560 GPU.
Memory usage by ffmpeg increases steadily and is not released, even when load is stable.
Within a few hours, all 24 GB of system RAM is consumed, causing system instability or a complete freeze.
Switching to an NVIDIA GPU resolves the issue entirely, suggesting the root cause may be related to AMD hardware acceleration (VAAPI/AMF), driver compatibility, or how Frigate integrates with FFmpeg.
Temporary Workaround:
Currently, the system is running in CPU-only mode with the RX560 removed. This avoids the issue but results in higher CPU usage and is not sustainable long term.
Additional Notes:
The issue appears to be triggered more aggressively by certain low-cost ONVIF-compatible Chinese cameras.
Frigate was used with its default hardware acceleration behavior — no manual overrides or custom ffmpeg arguments.
Steps to reproduce
...
Version
0.15.0 (stable9
In which browser(s) are you experiencing the issue with?
Vivaldi, Chrome, edge, and Opera GX
Frigate config file
docker-compose file or Docker CLI command
Relevant Frigate log output
Relevant go2rtc log output
Operating system
Other Linux
Install method
Docker Compose
Network connection
Wired
Camera make and model
hikvision, and chinese generic rtsp cameras
Screenshots of the Frigate UI's System metrics pages
I plan to reinstall the full setup with the RX560 next week in order to collect detailed logs and error messages. These will be added to this report to help further isolate the issue.
Any other information that may be helpful
When I disable the camera that is currently causing the memory spike, the issue shifts to another camera at random. However, the memory leak never occurs with more than one camera at a time — it seems to affect only a single stream at any given moment, and the affected stream varies unpredictably.
Beta Was this translation helpful? Give feedback.
All reactions