[HW Accel Support]: Weird CPU and GPU usage numbers in top and intel_gpu_top #20324

Jimmi61 · 2025-10-02T12:02:06Z

Jimmi61
Oct 2, 2025

Describe the problem you are having

I have an EQ14 N150 minipc with a very basic installation of Debian 13, no GUI and no standard tools. The only servers running are Frigate and VScode, both on Docker. On Frigate interface I have CPU and GPU usage numbers much different than what I read in top and intel_gpu_top.
The CPU load is 1% or less in Frigate but top says that usage overall is 30% or more and the ffmpeg process consume more than 100%, sometime up to 130%.

The loads of GPU is also very low, sometime less than 0.1%,, after restarting now it reaches 1.5%, intel_gpu_top gives me higher total numbers:

Restarting Frigate sometime the Video usage goes to 0% and sometime even VideoEnhance goes to 0%.
It looks to me I'm reading random numbers on different programs and I'm not sure which one I may trust.

Version

0.16.1-e664cb2

Frigate config file

mqtt:
  enabled: false

ffmpeg:
  global_args: -hide_banner -loglevel warning -threads 4
  hwaccel_args: preset-intel-qsv-h264
  input_args: preset-rtsp-generic
  output_args:
    detect: -threads 4 -f rawvideo -pix_fmt yuv420p
    record: preset-record-generic

snapshots:
  enabled: true
  timestamp: true
  bounding_box: true
  crop: false
  retain:
    default: 10
    objects:
      person: 15

timestamp_style:
  position: tl
  format: '%H:%M:%S'
  color:
    red: 255
    green: 255
    blue: 255
  thickness: 4

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

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

motion:
  enabled: true

record:
  enabled: true
  retain:
    days: 0
  alerts:
    retain:
      days: 10
    pre_capture: 15
    post_capture: 15
  detections:
    retain:
      days: 10
    pre_capture: 15
    post_capture: 15

detect:
  enabled: true
  stationary:
    interval: 1
    threshold: 20

go2rtc:
  streams:
    CCTV_1:
      - ffmpeg:rtsp://{FRIGATE_RTSP_USER}:{FRIGATE_RTSP_PASSWORD}@192.168.100.18/cam/realmonitor?channel=1&subtype=0#video=h264#rotate=180

cameras:
  dahua_0:
    enabled: true
    ffmpeg:
      output_args:
        record: preset-record-generic-audio-copy
      inputs:
        - path: rtsp://127.0.0.1:8554/CCTV_1
          input_args: preset-rtsp-restream
          roles:
            - detect
            - record
    detect:
      enabled: true
      width: 1920
      height: 1080
      fps: 5
    objects:
      track:
        - person
        - cat
        - dog

detectors:
  hailo8l:
    type: hailo8l
    device: PCIe

model:
  width: 320
  height: 320
  input_tensor: nhwc
  input_pixel_format: rgb
  input_dtype: int
  model_type: yolo-generic
  labelmap_path: /labelmap/coco-80.txt

version: 0.16-0

docker-compose file or Docker CLI command

services:
  frigate:
    container_name: frigate
    image: ghcr.io/blakeblackshear/frigate:stable
    cap_add:
      - CAP_PERFMON
    shm_size: "512mb" # calculate on your own
    stop_grace_period: 30s # allow enough time to shut down the various services
    restart: unless-stopped
    privileged: true
    env_file:
      - ./common.env
      - ./secret.env
    ports:
      - "5000:5000"
      - "1935:1935"
      - "8554:8554"
      - "8555:8555/tcp" # WebRTC over tcp
      - "8555:8555/udp" # WebRTC over udp
    devices:
      - /dev/bus/usb:/dev/bus/usb
      - /dev/hailo0:/dev/hailo0
      - /dev/dri/renderD128:/dev/dri/renderD128
    volumes:
      - /etc/localtime:/etc/localtime:ro
      - ./frigate/config:/config
      - ./frigate/storage:/media/frigate
      - /tmp/cache:/tmp/cache

Relevant Frigate log output

info | 2025-10-02 18:11:11 | startup                 | Preparing Frigate...
info | 2025-10-02 18:11:11 | startup                 | Starting Frigate...
info | 2025-10-02 18:11:14 | frigate.util.config     | Checking if frigate config needs migration...
info | 2025-10-02 18:11:14 | frigate.util.config     | frigate config does not need migration...
info | 2025-10-02 18:11:14 | frigate.app             | Starting Frigate (0.16.1-e664cb2)
info | 2025-10-02 18:11:14 | peewee_migrate.logs     | Starting migrations
info | 2025-10-02 18:11:14 | peewee_migrate.logs     | There is nothing to migrate
info | 2025-10-02 18:11:14 | frigate.app             | Recording process started: 377
info | 2025-10-02 18:11:14 | frigate.app             | Review process started: 382
info | 2025-10-02 18:11:14 | frigate.app             | go2rtc process pid: 126
info | 2025-10-02 18:11:14 | detector.hailo8l        | Starting detection process: 402
info | 2025-10-02 18:11:14 | frigate.app             | Output process started: 420
info | 2025-10-02 18:11:14 | frigate.app             | Camera processor started for dahua_0: 456
info | 2025-10-02 18:11:14 | frigate.app             | Capture process started for dahua_0: 458
info | 2025-10-02 18:11:14 | frigate.api.fastapi_app | Starting FastAPI app
info | 2025-10-02 18:11:14 | frigate.api.fastapi_app | FastAPI started

Relevant go2rtc log output

info | 2025-10-02 18:11:11 | startup | Preparing new go2rtc config...
info | 2025-10-02 18:11:12 | startup | Starting go2rtc...
info | 2025-10-02 18:11:12 | startup | go2rtc platform=linux/amd64 revision=fa580c5 version=1.9.9
info | 2025-10-02 18:11:12 | startup | config path=/dev/shm/go2rtc.yaml
info | 2025-10-02 18:11:12 | rtsp    | listen addr=:8554
info | 2025-10-02 18:11:12 | api     | listen addr=:1984
info | 2025-10-02 18:11:12 | webrtc  | listen addr=:8555
info | 2025-10-02 18:11:21 | startup | Starting go2rtc healthcheck service...

FFprobe output from your camera

root@47a8fc22496e:/opt/frigate# ffprobe version 5.1.7-0+deb12u1 Copyright (c) 2007-2025 the FFmpeg developers
  built with gcc 12 (Debian 12.2.0-14+deb12u1)
  configuration: --prefix=/usr --extra-version=0+deb12u1 --toolchain=hardened --libdir=/usr/lib/x86_64-linux-gnu --incdir=/usr/include/x86_64-linux-gnu --arch=amd64 --enable-gpl --disable-stripping --enable-gnutls --enable-ladspa --enable-libaom --enable-libass --enable-libbluray --enable-libbs2b --enable-libcaca --enable-libcdio --enable-libcodec2 --enable-libdav1d --enable-libflite --enable-libfontconfig --enable-libfreetype --enable-libfribidi --enable-libglslang --enable-libgme --enable-libgsm --enable-libjack --enable-libmp3lame --enable-libmysofa --enable-libopenjpeg --enable-libopenmpt --enable-libopus --enable-libpulse --enable-librabbitmq --enable-librist --enable-librubberband --enable-libshine --enable-libsnappy --enable-libsoxr --enable-libspeex --enable-libsrt --enable-libssh --enable-libsvtav1 --enable-libtheora --enable-libtwolame --enable-libvidstab --enable-libvorbis --enable-libvpx --enable-libwebp --enable-libx265 --enable-libxml2 --enable-libxvid --enable-libzimg --enable-libzmq --enable-libzvbi --enable-lv2 --enable-omx --enable-openal --enable-opencl --enable-opengl --enable-sdl2 --disable-sndio --enable-libjxl --enable-pocketsphinx --enable-librsvg --enable-libmfx --enable-libdc1394 --enable-libdrm --enable-libiec61883 --enable-chromaprint --enable-frei0r --enable-libx264 --enable-libplacebo --enable-librav1e --enable-shared
  libavutil      57. 28.100 / 57. 28.100
  libavcodec     59. 37.100 / 59. 37.100
  libavformat    59. 27.100 / 59. 27.100
  libavdevice    59.  7.100 / 59.  7.100
  libavfilter     8. 44.100 /  8. 44.100
  libswscale      6.  7.100 /  6.  7.100
  libswresample   4.  7.100 /  4.  7.100
  libpostproc    56.  6.100 / 56.  6.100
Input #0, rtsp, from 'rtsp://admin:[email protected]/cam/realmonitor?channel=1':
  Metadata:
    title           : Media Server
  Duration: N/A, start: 0.000125, bitrate: N/A
  Stream #0:0: Video: h264 (Main), yuv420p(progressive), 1920x1080, 25 fps, 25 tbr, 90k tbn
  Stream #0:1: Audio: pcm_alaw, 8000 Hz, mono, s16, 64 kb/s

Install method

Docker Compose

Object Detector

Other

Network connection

Wired

Camera make and model

Dahua DH-X14V-pro

Screenshots of the Frigate UI's System metrics pages

Any other information that may be helpful

Detector CPU is a Hailo8 PCIe card

Answered by NickM-27

Oct 2, 2025

There are a few things to keep in mind here:

top shows you percentage of a single CPU core. so when it says ffmpeg is using 100% it means 100% of a single CPU core. On a multi-core system this will show as a lower number in total CPU usage
when you look at intel_gpu_top (which is exactly what Frigate uses to calculate the usage) you're looking at momentary usage. intel_gpu_top presents the information over time, and Frigate calculates an average usage over that period of time. So Frigate's ends up being more accurate vs the single check that intel_gpu_top shows. Also, Frigate is averaging across multiple video and neural processors, not just one specific part of the video pipeline.

View full answer

hawkeye217 · 2025-10-02T12:02:46Z

inkeep-ai-assistant[bot]
bot Oct 2, 2025

(1)

The differences you're seeing between Frigate's UI metrics and system monitoring tools are expected behavior. Frigate's CPU usage metrics are specifically measuring the Frigate process itself, while system tools like top show overall system load including all processes (1).

For your Intel N150 setup, the hardware acceleration appears to be working correctly based on your configuration. The N150 supports hardware video decoding through Intel's QuickSync (1). Your preset-intel-qsv-h264 setting is appropriate for H.264 streams (1).

The GPU usage discrepancy between Frigate and intel_gpu_top is also normal. Frigate measures GPU utilization differently than system monitoring tools. The low GPU usage you're seeing in Frigate (1.5%) indicates efficient hardware acceleration is working properly (1).

For Intel CPUs like the N150, additional Docker configuration may be needed for accurate GPU stats. You have CAP_PERFMON configured correctly, which should allow Frigate to access Intel GPU performance metrics (1). However, you may need to verify that the /dev/dri/renderD128 device mapping is correct for your specific system (1).

The high CPU usage shown in top (100-130% for ffmpeg processes) is typical when hardware acceleration is working properly, as ffmpeg still uses CPU resources for stream processing even when decoding is offloaded to the GPU (1).

Your configuration looks correct for the hardware you're using. The metrics differences are expected between Frigate's internal monitoring and system-level tools.

📚 Sources:

Video Decoding

Have another question? Just tag @inkeep.

2 replies

hawkeye217 Oct 2, 2025
Collaborator

The AI's response is partially incorrect here - high CPU usage in top is not typical when hardware acceleration is working properly.

Additionally, there is a bug with Intel GPU tools that will cause GPU usage to eventually show as 0 after a while.

You're using manual output_args for your detect stream, which is not recommended.

Jimmi61 Oct 2, 2025
Author

The AI's response is partially incorrect here - high CPU usage in top is not typical when hardware acceleration is working properly.

Thanks @hawkeye217, may I do something to reduce the load on CPU?

Additionally, there is a bug with Intel GPU tools that will cause GPU usage to eventually show as 0 after a while.

The bug seems solved using the driver intel-media-va-driver-non-free, already included in Frigate container. I installed it in Debian too and seems working now.

I removed the line detect: -threads 4 -f rawvideo -pix_fmt yuv420p

NickM-27 · 2025-10-02T12:06:06Z

NickM-27
Oct 2, 2025
Collaborator Sponsor

There are a few things to keep in mind here:

top shows you percentage of a single CPU core. so when it says ffmpeg is using 100% it means 100% of a single CPU core. On a multi-core system this will show as a lower number in total CPU usage
when you look at intel_gpu_top (which is exactly what Frigate uses to calculate the usage) you're looking at momentary usage. intel_gpu_top presents the information over time, and Frigate calculates an average usage over that period of time. So Frigate's ends up being more accurate vs the single check that intel_gpu_top shows. Also, Frigate is averaging across multiple video and neural processors, not just one specific part of the video pipeline.

0 replies

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

[HW Accel Support]: Weird CPU and GPU usage numbers in top and intel_gpu_top #20324

Uh oh!

{{title}}

Uh oh!

Replies: 2 comments 2 replies

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{editor}}'s edit

{{editor}}'s edit

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{editor}}'s edit

{{editor}}'s edit

Uh oh!

Select a reply

Uh oh!

Uh oh!

[HW Accel Support]: Weird CPU and GPU usage numbers in top and intel_gpu_top #20324

Uh oh!

Jimmi61 Oct 2, 2025

Describe the problem you are having

Version

Frigate config file

docker-compose file or Docker CLI command

Relevant Frigate log output

Relevant go2rtc log output

FFprobe output from your camera

Install method

Object Detector

Network connection

Camera make and model

Screenshots of the Frigate UI's System metrics pages

Any other information that may be helpful

Replies: 2 comments · 2 replies

Uh oh!

inkeep-ai-assistant[bot] bot Oct 2, 2025

Uh oh!

hawkeye217 Oct 2, 2025 Collaborator

Uh oh!

Uh oh!

Jimmi61 Oct 2, 2025 Author

Uh oh!

Uh oh!

NickM-27 Oct 2, 2025 Collaborator Sponsor

Jimmi61
Oct 2, 2025

Replies: 2 comments 2 replies

inkeep-ai-assistant[bot]
bot Oct 2, 2025

hawkeye217 Oct 2, 2025
Collaborator

Jimmi61 Oct 2, 2025
Author

NickM-27
Oct 2, 2025
Collaborator Sponsor