[HW Accel Support]: Heavy CPU usage after upgrade from 13.2 to 14.0 beta #12614

edkoch25 · 2024-07-25T18:27:11Z

edkoch25
Jul 25, 2024

Describe the problem you are having

My cpu usage has spiked from an average of 12 to over 30 with spikes to 75 % (when detection kicks in). What if anything changed and any recommendations on how to improve cpu usage?

I'm running on a Qnap with TS-664 Intel Celeron N5105 / N5095 quad-core processor.

Version

0.14.0 Beta

Frigate config file

mqtt:
  host: 192.168.5.150
  port: 1883
  user: frigate
  password: 

ffmpeg:
  hwaccel_args: preset-vaapi

birdseye:
  enabled: true
  width: 3840
  height: 2160
  quality: 10
  mode: continuous

cameras:
  front_yard:
    ffmpeg:
      inputs:
        - path: rtsp://admin:[email protected]:554/cam/realmonitor?channel=1&subtype=1
          roles:
            - detect
        - path: rtsp://admin:[email protected]:554/
          roles:
            - record
  garage_door:
    ffmpeg:
      inputs:
        - path: rtsp://admin:[email protected]:554/cam/realmonitor?channel=1&subtype=1
          roles:
            - detect
        - path: rtsp://admin:[email protected]:554/
          roles:
            - record
  back_patio:
    ffmpeg:
      inputs:
        - path: rtsp://admin:[email protected]:554/
          roles:
            - record
  side_yard:
    ffmpeg:
      inputs:
        - path: rtsp://admin:[email protected]:554/cam/realmonitor?channel=1&subtype=1
          roles:
            - detect
        - path: rtsp://admin:[email protected]:554/
          roles:
            - record

record:
  enabled: true
  retain:
    days: 13
    mode: all

objects:
  track:
    - person

detectors:
  coral:
    type: edgetpu
    device: pci:0
    detect:
      width: 704
      height: 480
      fps: 15

snapshots:
  enabled: true
  timestamp: false
  bounding_box: true
  retain:
    default: 10
version: 0.14

docker-compose file or Docker CLI command

services:
  frigate:
    image: ghcr.io/blakeblackshear/frigate:0.14.0-beta4
    container_name: frigate
    restart: unless-stopped
    privileged: true
    shm_size: "512mb"
    environment:
      - FRIGATE_MQTT_HOST=192.168.5.150
      - FRIGATE_MQTT_PORT=1883
      - FRIGATE_MQTT_USER=frigate
      - FRIGATE_MQTT_PASSWORD=password
      - FRIGATE_FFMPEG_HWACCEL_ARGS="-hwaccel vaapi -hwaccel_device /dev/dri/renderD128"  
      - LIBVA_DRIVER_NAME=iHD
    volumes:
      - /share/CACHEDEV3_DATA/frigate/config:/config
      - /dev:/dev
      - /etc/localtime:/etc/localtime
      - /share/CACHEDEV3_DATA/frigate:/media
      - type: tmpfs
        target: /tmp/cache
        tmpfs:
          size: 1000000000  # 1GB of memory
    devices:
      - /dev/bus/pci:/dev/bus/pci
      - /dev/dri:/dev/dri
      - /dev/dri/renderD128:/dev/dri/renderD128
    ports:
      - "5000:5000"
      - "1935:1935"

Relevant log output

2024-07-25 14:14:12.163300418  [INFO] Preparing Frigate...
2024-07-25 14:14:12.196188029  [INFO] Starting Frigate...
2024-07-25 14:14:15.391962406  [2024-07-25 14:14:15] frigate.app                    INFO    : Starting Frigate (0.14.0-a4eb435)
2024-07-25 14:14:15.392165137  [2024-07-25 14:14:15] frigate.util.config            INFO    : Checking if frigate config needs migration...
2024-07-25 14:14:15.420883127  [2024-07-25 14:14:15] frigate.util.config            INFO    : frigate config does not need migration...
2024-07-25 14:14:22.596914666  [2024-07-25 14:14:22] peewee_migrate.logs            INFO    : Starting migrations
2024-07-25 14:14:22.597917523  [2024-07-25 14:14:22] peewee_migrate.logs            INFO    : There is nothing to migrate
2024-07-25 14:14:22.871734253  [2024-07-25 14:14:22] frigate.app                    INFO    : Recording process started: 361
2024-07-25 14:14:22.878432061  [2024-07-25 14:14:22] frigate.app                    INFO    : Recording process started: 367
2024-07-25 14:14:22.881013789  [2024-07-25 14:14:22] frigate.app                    INFO    : go2rtc process pid: 110
2024-07-25 14:14:22.920351695  [2024-07-25 14:14:22] frigate.app                    INFO    : Output process started: 392
2024-07-25 14:14:22.938780993  [2024-07-25 14:14:22] detector.coral                 INFO    : Starting detection process: 390
2024-07-25 14:14:22.943403972  [2024-07-25 14:14:22] frigate.detectors.plugins.edgetpu_tfl INFO    : Attempting to load TPU as pci:0
2024-07-25 14:14:22.943408722  [2024-07-25 14:14:22] frigate.detectors.plugins.edgetpu_tfl INFO    : TPU found
2024-07-25 14:14:23.436219101  [2024-07-25 14:14:23] frigate.app                    INFO    : Camera processor started for front_yard: 448
2024-07-25 14:14:23.443136483  [2024-07-25 14:14:23] frigate.app                    INFO    : Camera processor started for garage_door: 454
2024-07-25 14:14:23.449402746  [2024-07-25 14:14:23] frigate.app                    INFO    : Camera processor started for back_patio: 459
2024-07-25 14:14:23.455047307  [2024-07-25 14:14:23] frigate.app                    INFO    : Camera processor started for side_yard: 465
2024-07-25 14:14:23.460902106  [2024-07-25 14:14:23] frigate.app                    INFO    : Capture process started for front_yard: 468
2024-07-25 14:14:23.466684306  [2024-07-25 14:14:23] frigate.app                    INFO    : Capture process started for garage_door: 472
2024-07-25 14:14:23.472598055  [2024-07-25 14:14:23] frigate.app                    INFO    : Capture process started for back_patio: 474
2024-07-25 14:14:23.479517586  [2024-07-25 14:14:23] frigate.app                    INFO    : Capture process started for side_yard: 479

FFprobe output from your camera

don't have it installed right now.

Operating system

Other Linux

Install method

Docker Compose

Network connection

Wired

Camera make and model

Dahua Starlight 8MP 4K Outdoor PoE IP Security Dome Camera IPC-HDBW2831R-ZAS-S2

Any other information that may be helpful

No response

Answered by edkoch25

Jul 25, 2024

update: think the back patio detect setting was causing the issue. I'm still seeing a few spikes into the 20's but it's holding pretty steady at 12 %.

thanks again hawkeye (excellent name choice)!

View full answer

NickM-27 · 2024-07-25T18:28:58Z

NickM-27
Jul 25, 2024
Collaborator Sponsor

Would be good to see screenshots of the system and cameras metrics pages

also,

    detect:
      width: 704
      height: 480
      fps: 15

under detectors is doing nothing, I would suggest removing it

2 replies

edkoch25 Jul 25, 2024
Author

removed, ty.

+screen grabs

NickM-27 Jul 25, 2024
Collaborator Sponsor

Not seeing anything that sticks out as high CPU usage, the birdseye resolution being 4K will definitely use a decent amount of CPU but besides that I don't see anything that stands out as an issue

edkoch25 · 2024-07-25T20:23:50Z

edkoch25
Jul 25, 2024
Author

Thanks Nick, something changed between 13.2 and 14.0 beta 4.

Is there a different hardware acceleration type I should try? Tried setting LIBVA_DRIVER_NAME=i965 in my docker...frigate came up but camera feeds were dead. guessing driver compatibility issue with camera type?

0 replies

hawkeye217 · 2024-07-25T20:40:50Z

hawkeye217
Jul 25, 2024
Collaborator

Your back_patio camera doesn't have a detect role defined. The cpu usage for that camera is pretty high as a result. Did you forget to set up a detect role input stream for it when you migrated to the new version?

0 replies

edkoch25 · 2024-07-25T20:46:20Z

edkoch25
Jul 25, 2024
Author

Nice catch, ty.

I had been leaving the detect portion out for that camera because I didn't require or want it and turning it off at the camera setting wasn't sticking. will reconfigure and see if that helps.

3 replies

hawkeye217 Jul 25, 2024
Collaborator

When you don't have a detect role defined, it will default to the single input stream you've set. A high-res stream has to be decoded and reencoded for birdseye. Since detection for that camera is not disabled either (you can permanently disable it with enabled: False in the detect: section of that camera, see the reference config), object detection will be run on high-res frames too.

All of those things would certainly contribute to higher cpu load.

edkoch25 Jul 25, 2024
Author

update: think the back patio detect setting was causing the issue. I'm still seeing a few spikes into the 20's but it's holding pretty steady at 12 %.

thanks again hawkeye (excellent name choice)!

Answer selected by NickM-27

hawkeye217 Jul 25, 2024
Collaborator

Great! Happy to help.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[HW Accel Support]: Heavy CPU usage after upgrade from 13.2 to 14.0 beta #12614

{{title}}

{{editor}}'s edit

{{editor}}'s edit

Replies: 4 comments 5 replies

{{title}}

{{title}}

{{title}}

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

Select a reply

[HW Accel Support]: Heavy CPU usage after upgrade from 13.2 to 14.0 beta #12614

edkoch25 Jul 25, 2024

Describe the problem you are having

Version

Frigate config file

docker-compose file or Docker CLI command

Relevant log output

FFprobe output from your camera

Operating system

Install method

Network connection

Camera make and model

Any other information that may be helpful

Replies: 4 comments · 5 replies

NickM-27 Jul 25, 2024 Collaborator Sponsor

edkoch25 Jul 25, 2024 Author

NickM-27 Jul 25, 2024 Collaborator Sponsor

edkoch25 Jul 25, 2024 Author

hawkeye217 Jul 25, 2024 Collaborator

edkoch25 Jul 25, 2024 Author

hawkeye217 Jul 25, 2024 Collaborator

edkoch25 Jul 25, 2024 Author

hawkeye217 Jul 25, 2024 Collaborator

edkoch25
Jul 25, 2024

Replies: 4 comments 5 replies

NickM-27
Jul 25, 2024
Collaborator Sponsor

edkoch25 Jul 25, 2024
Author

NickM-27 Jul 25, 2024
Collaborator Sponsor

edkoch25
Jul 25, 2024
Author

hawkeye217
Jul 25, 2024
Collaborator

edkoch25
Jul 25, 2024
Author

hawkeye217 Jul 25, 2024
Collaborator

edkoch25 Jul 25, 2024
Author

hawkeye217 Jul 25, 2024
Collaborator