[Config Support]: High FFMPEG CPU Usage (93%)

[szoller]

Describe the problem you are having

Hello,

i experience high CPU load by frigate even when using a Google Coral TPU (USB) which seems to be successfully connected.

Frigate runs as docker container on my Synology 918+ (16GB RAM).

I was told that using a Coral TPU it would not take much ressources, but it does. My Synology has around 35-40% CPU load, before it was like <10%.

Maybe i just have some wrong settings, so I attached my config below.

I'm just running one single camera with maybe a second one later. Just FullHD.

Detector CPU Usage is around 4% - at least thats shown on metrics page

So is there something I can improve?

Version

0.15.0-6cb5cfb

Frigate config file

database:
  path: /config/frigate.db
  
mqtt:
  enabled: false

ffmpeg:
  hwaccel_args: auto

cameras:
  Vorgarten: # <------ Name the camera
    enabled: true
    ffmpeg:
      inputs:
        - path: rtsp://admin:[email protected]:554/H.264 # <----- The stream you want to use for detection
          roles:
            - detect
    detect:
      enabled: true # <---- disable detection until you have a working camera feed
      width: 2560
      height: 1440
    zones:
      Parkbereich:
        coordinates: 0.111,0.577,0,0.585,0.039,0.288,0.125,0.305
        loitering_time: 0
      Haustür:
        coordinates: 0.626,0.605,0.633,0.123,0.496,0.169,0.498,0.48,0.514,0.599
        loitering_time: 0
    review:
      alerts:
        required_zones:
          - Parkbereich
          - Haustür
record:
  enabled: True
  retain:
    days: 1
    mode: motion
  alerts:
    retain:
      days: 7
      mode: motion
  detections:
    retain:
      days: 7
      mode: motion

objects:
  track:
    - person
    - cat
    - dog
    - car

review:
  alerts:
    labels:
      - car
      - person
  detections:
    labels:
      - car
      - cat
      - dog
      - person

detectors:
  coral:
    type: edgetpu
    device: usb
version: 0.15-1
semantic_search:
  enabled: true
  reindex: false
  model_size: small

Relevant Frigate log output

info    | 2025-06-12 15:09:18 | frigate.api.fastapi_app               | FastAPI started
info    | 2025-06-12 15:09:20 | frigate.detectors.plugins.edgetpu_tfl | TPU found
error   | 2025-06-12 15:09:43 | frigate.util.services                 | Unable to poll intel GPU stats: Failed to initialize PMU! (Invalid argument)
unknown | 2025-06-12 15:29:49 | unknown                               | 25-06-12 15:09:43.481635755
info    | 2025-06-12 15:18:49 | frigate.comms.dispatcher              | Setting motion threshold for Vorgarten: 30
info    | 2025-06-12 15:18:49 | frigate.comms.dispatcher              | Setting motion contour area for Vorgarten: 10

Relevant go2rtc log output

none

Frigate stats

No response

Operating system

Other

Install method

Docker Compose

docker-compose file or Docker CLI command

version: "3.9"
services:
  frigate:
    container_name: frigate
    privileged: true # this may not be necessary for all setups
    restart: unless-stopped
    stop_grace_period: 30s # allow enough time to shut down the various services
    image: ghcr.io/blakeblackshear/frigate:stable
    shm_size: "512mb" # update for your cameras based on calculation above
    devices:
      - /dev/bus/usb:/dev/bus/usb # Passes the USB Coral, needs to be modified for other versions
     # - /dev/apex_0:/dev/apex_0 # Passes a PCIe Coral, follow driver instructions here https://coral.ai/docs/m2/get-started/#2a-on-linux
     # - /dev/video11:/dev/video11 # For Raspberry Pi 4B
     # - /dev/dri/renderD128:/dev/dri/renderD128 # For intel hwaccel, needs to be updated for your hardware
    volumes:
      - /etc/localtime:/etc/localtime:ro
      - /volume1/docker/frigate/config:/config
      - /volume1/surveillance:/media/frigate
      - type: tmpfs # Optional: 1GB of memory, reduces SSD/SD Card wear
        target: /tmp/cache
        tmpfs:
          size: 1000000000
    ports:
      - "8971:8971"
      - "XYXY:5000" # Internal unauthenticated access. Expose carefully.
      - "8554:8554" # RTSP feeds
      - "8555:8555/tcp" # WebRTC over tcp
      - "8555:8555/udp" # WebRTC over udp
    environment:
      FRIGATE_RTSP_PASSWORD: "XYXYXY"

Object Detector

Coral

Screenshots of the Frigate UI's System metrics pages

No response

Any other information that may be helpful

No response

[hawkeye217]

You should enable hardware acceleration (I would try preset-vaapi first) and choose a lower detect resolution.

[szoller]

So I changed those parts:
ffmpeg:
hwaccel_args: preset-vaapi

detect:
enabled: true # <---- disable detection until you have a working camera feed
width: 1920
height: 1080

(from 2K to FullHD should help a bit)

Am I right?

edit: GPU seems not to work:
Tag frigate.util.services
Message Unable to poll intel GPU stats: Failed to initialize PMU! (Invalid argument)

[hawkeye217]

That looks good. The error message is not necessarily indicative of the GPU itself not functioning, just that Frigate can't fetch the stats.

What is your CPU usage now?

[pipip]

you also need to change your docker compose file and "activate" your GPU in the device section

[szoller]

On the bottom left ("CPU") i'm shown values between 23 and 75%, on the right i see "Camera has high FFMPEG CPU usage (54%)... still too much i guess

[hawkeye217]

It would be good to see screenshots of your system and camera metrics pages.

[hawkeye217]

Also, what CPU does your Synology have?