[HW Accel Support]: Intel N100 CPU with Home Assistant OS 14.2

[kkevindev]

Describe the problem you are having

Just does not seem to be using the GPU on my machine. Having a high CPU % (± 20% per camera) and 0% GPU.

Also, weird thing I've noticed is that I've disabled my detect in the config but it says something about detecting started

Version

0.15-1

Frigate config file

version: 0.15-1

go2rtc:
  streams:
    poort:
      - rtsp://admin:[email protected]/cam/realmonitor?channel=1&subtype=0

birdseye:
  enabled: false

ui:
  time_format: 24hour

environment_vars:
  LIBVA_DRIVER_NAME: iHD

record:
  enabled: true
  retain:
    mode: all
    days: 7

snapshots:
  enabled: false

detect:
  enabled: false

ffmpeg:
  hwaccel_args: preset-vaapi

cameras:
  poort:
    enabled: true
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/poort
          roles:
            - record

docker-compose file or Docker CLI command

-

Relevant Frigate log output

Has some "[vf#1:0 @ 0x55ff29bf9900] Task finished with error code: -5 (Input/output error)" errors sometimes.

Relevant go2rtc log output

no errors

FFprobe output from your camera

-

Install method

HassOS Addon

Object Detector

Other

Network connection

Wireless

Camera make and model

Imou Rex 3D

Screenshots of the Frigate UI's System metrics pages

Any other information that may be helpful

No response

[hawkeye217]

Even though you have disabled detection, Frigate still needs to decode a stream for other features. See the note in the official docs about the detect role: https://docs.frigate.video/configuration/cameras/#setting-up-camera-inputs

If you only define one stream in your inputs and do not assign a detect role to it, Frigate will automatically assign it the detect role. Frigate will always decode a stream to support motion detection, Birdseye, the API image endpoints, and other features, even if you have disabled object detection with enabled: False in your config's detect section.

If you plan to use Frigate for recording only, it is still recommended to define a detect role for a low resolution stream to minimize resource usage from the required stream decoding.

[kkevindev]

Thanks, best thing for me to do to virtually disable it with a resolution of 1x1 for the detect role?

define a detect role for a low resolution stream to minimize resource usage from the required stream decoding.

[NickM-27]

You should be using qsv not vaapi with this hardware. We can tell it is using the GPU though based on the errors. There is a bug with Intel GPU stats for some users.

[kkevindev]

@NickM-27 2880x1620

[NickM-27]

I'd suggest lowering the detect size then, that is a lot of detail to process without good reason. Your camera resolution and detect width / height do not need to match

[kkevindev]

@NickM-27 I have set:

detect:
  enabled: false

model:
  width: 1
  height: 1

but this does not decrease the CPU % or increase the GPU %.

[NickM-27]

that is the wrong height / width to set. You need to change detect width / height

[kkevindev]

Doesn't change the CPU % unless I take it all the way back to 1 (like the example above) but then my playback in the homescreen is F-ed.

[Imwithstupid007]

Not trying to steal this thread......
There is a bug with Intel GPU stats for some users.
Is this bug documented somewhere?

[kkevindev]

When i tried hwaccel_args: preset-intel-qsv-h265 #16609 (reply in thread) my CPU % was the same with or without that line. So even if I would have that bug, howcome the CPU % would stay the same? Would think that the CPU % would drop if the GPU would actually be used.

[Imwithstupid007]

I also have an Alderlake-n generation cpu (Odroid H4 +ultra, i3-N305).
What i'm seeing is that after a fresh restart of the NAS the iGPU shows activity (via Intel_gpu_top) which is also visible in the metrics page of Frigate. Then, after a few hours to a few days, something happens (I don't know what), but the activity in the Frigate metrics page shows always zero. Intel_gpu_top is also different then. The activity bars are constant zero but you can see activity in the top (Mhz/RC6 and irq/s).
The above behavior doesn't seem to impact the CPU as its activity stays the same. So the iGPU seems to do something. Maybe it is only a visual bug.

Restarting the compose container or docker completely does not result in any change. Only a complete reboot of the NAS brings things back to the expected.

[Summit1423]

Полный перезапуск контейнера Compose или Docker не приводит ни к каким изменениям. Только полная перезагрузка NAS возвращает все к ожидаемому.

Have you solved this problem?

[Imwithstupid007]

@Summit1423 ; I'm not sure it is a real problem. I experimented with an extra camera to increase the load on the GPU. As a result the GPU activity (metrics page) was visible again (I mean activities different from zero). Deleting the extra camera and it drop down to zero again.

I have the impression that it is some sort of cosmetical bug. After some time, while the GPU is under a certain load the GPU activity shows all zero values.

[xconverge]

Cross posting to this thread but I am testing a kernel patch at the moment because I also started to wonder more about this mysterious bug with the intel gpu stats

#16438 (comment)

I will update there (and here probably) if my testing is conclusive or inconclusive