[HW Accel Support]: Asus NUC 14 Pro / Intel Core Ultra 7 155H

[matknowles]

Describe the problem you are having

Hello, Can't seem to get hardware acceleration working with my new Asus Pro NUC14RVHU7 NUC. Initially tried just vaapi but got no frames, so tried all combinations of environment_vars: LIBVA_DRIVER_NAME: i965 + ffmpeg: hwaccel_args: preset-vaapi and hwaccel_args: preset-intel-qsv-h265 (am using all H265 cameras). Frigate is working if I comment out the ffmpeg hwaccel_args at about 26% CPU, but isn't ideal. I can only find that the Core 7 Ultra 155H uses "Intel Arc" graphics.

Trying qsv gives me errors like:
Ffmpeg process crashed unexpectedly for FrontPath. 2024-11-23 18:16:07.540300350 [2024-11-23 18:16:07] watchdog.FrontPath ERROR : The following ffmpeg logs include the last 100 lines prior to exit. 2024-11-23 18:16:07.540360396 [2024-11-23 18:16:07] ffmpeg.FrontPath.detect ERROR : [AVHWDeviceContext @ 0x55d77a47b680] No VA display found for device -hwaccel_output_format. 2024-11-23 18:16:07.540421443 [2024-11-23 18:16:07] ffmpeg.FrontPath.detect ERROR : Device creation failed: -22. 2024-11-23 18:16:07.540460040 [2024-11-23 18:16:07] ffmpeg.FrontPath.detect ERROR : Failed to set value '-hwaccel_output_format' for option 'qsv_device': Invalid argument 2024-11-23 18:16:07.540500334 [2024-11-23 18:16:07] ffmpeg.FrontPath.detect ERROR : Error parsing global options: Invalid argument

Using environment vars i965 and vaapi gives:
[2024-11-23 19:25:39] frigate.util.services ERROR : Unable to poll intel GPU stats: No device filter specified and no discrete/integrated i915 devices found 2024-11-23 19:25:39.325736919 2024-11-23 19:25:39.604099052 [2024-11-23 19:25:39] frigate.video ERROR : FrontPath: Unable to read frames from ffmpeg process.
and
[2024-11-23 19:25:58] ffmpeg.FrontPath.detect ERROR : Option hwaccel (use HW accelerated decoding) cannot be applied to output url vaapi -- you are trying to apply an input option to an output file or vice versa. Move this option before the file it belongs to. 2024-11-23 19:25:58.894778812 [2024-11-23 19:25:58] watchdog.FrontPath ERROR : The following ffmpeg logs include the last 100 lines prior to exit. 2024-11-23 19:25:58.894966237 [2024-11-23 19:25:58] ffmpeg.FrontPath.detect ERROR : Error parsing options for output file vaapi. 2024-11-23 19:25:58.894972777 [2024-11-23 19:25:58] ffmpeg.FrontPath.detect ERROR : Error opening output files: Invalid argument 2024-11-23 19:25:58.894979279 [2024-11-23 19:25:58] ffmpeg.FrontPath.detect ERROR : Option hwaccel (use HW accelerated decoding) cannot be applied to output url vaapi -- you are trying to apply an input option to an output file or vice versa. Move this option before the file it belongs to.

Just using vaapi:
[2024-11-23 19:30:28] frigate.video ERROR : FrontPath: Unable to read frames from ffmpeg process. 2024-11-23 19:30:28.648519630 [2024-11-23 19:30:28] frigate.video ERROR : FrontPath: ffmpeg process is not running. exiting capture thread... [2024-11-23 19:30:29] frigate.util.services ERROR : Unable to poll vainfo: b"error: XDG_RUNTIME_DIR is invalid or not set in the environment.\nerror: can't connect to X server!\nerror: failed to initialize display\n"

I haven't tried 0.15 beta yet, but I bought this new NUC to hopefully be powerful enough for Frigate going into the future. If I'm going to have a bad time I may as well return it.

It's just occured to me that I restored this home assistant install from a backup from my old NUC that died (hardware failure), so I wonder if the XDG_RUNTIME_DIR is invalid has anything to do with that? Apologies for my ignorance.

Version

0.14.1-f4f3cfa

Frigate config file

database:
  path: /config/frigate.db0

#environment_vars:
#  LIBVA_DRIVER_NAME: i965

logger:
  # Optional: default log level (default: shown below)
  default: info
  # Optional: module by module log level configuration
  logs:
    frigate.mqtt: error

model:
  path: plus://

mqtt:
  host: core-mosquitto
  user: mqtt
  password: 

#semantic_search:
#  enabled: true
#  reindex: true
#  model_size: large

#genai:
#  enabled: True
#  provider: gemini
#  api_key: ""
#  model: gemini-1.5-flash

ffmpeg: # try "ffmpeg -decoders | grep cuvid" in terminal to see available decoders
  #hwaccel_args: preset-vaapi # Intel H264+H265 (recommended as of 2023)
  #hwaccel_args: preset-intel-qsv-h265 #Intel >10th Gen H265 (use if problems with vaapi)
  #hwaccel_args: preset-intel-qsv-h264 #Intel >10th Gen H264
  #hwaccel_args: preset-rpi-64-h265 #Raspi H265
  output_args:
    record: preset-record-generic-audio-aac

detectors:
  coral:
    type: edgetpu
    device: usb

camera_groups:
  Birdseye:
    order: 1
    icon: LuBird
    cameras: birdseye

birdseye:
  # Optional: Enable birdseye view (default: shown below)
  enabled: true
  # Optional: Restream birdseye via RTSP (default: shown below)
  # NOTE: Enabling this will set birdseye to run 24/7 which may increase CPU usage somewhat.
  restream: true
  # Optional: Width of the output resolution (default: shown below)
  width: 1920
  # Optional: Height of the output resolution (default: shown below)
  height: 1080
  # Optional: Encoding quality of the mpeg1 feed (default: shown below)
  # 8 is default. 1 is the highest quality, and 31 is the lowest. Lower quality feeds utilize less CPU resources.
  quality: 4
  # Optional: Mode of the view. Available options are: objects, motion, and continuous
  #   objects - cameras are included if they have had a tracked object within the last 30 seconds
  #   motion - cameras are included if motion was detected in the last 30 seconds
  #   continuous - all cameras are included always
  mode: continuous
  # Optional: Threshold for camera activity to stop showing camera (default: shown below)
  #inactivity_threshold: 30
  # Optional: Configure the birdseye layout
  #layout:
    # Optional: Scaling factor for the layout calculator, range 1.0-5.0 (default: shown below)
   # scaling_factor: 2.0
    # Optional: Maximum number of cameras to show at one time, showing the most recent (default: show all cameras)
    #max_cameras: 1

go2rtc:
  streams:
    FrontPath:
      - rtsp://@192.168.1.11:554/main

detect:
  stationary:
    interval: 50
    threshold: 50

objects:
  track:
    - person
    - car
    - bicycle
    - motorcycle
    - bus
    - bird
    - cat
    - dog
    - horse
    - bear
    - hat
    - backpack
    - umbrella
    - eye glasses
    - handbag
    - suitcase
    - face
    - license_plate
    - amazon
    - ups
    - fedex
    - package
    #- dhl
    #- dpd
    #- fox
    #- squirrel
    #- waste_bin
  filters:
    #global filters
    # defaults below taken from Frigate+ documentation: https://docs.frigate.video/plus/
    person:
      min_score: 0.6 #0.56 calibrated for porch night
      threshold: 0.62 #0.63 calibrated for porch night
    car:
      min_score: 0.65
      threshold: 0.85
    bicycle:
      min_score: 0.5
      threshold: 0.7
    motorcycle:
      min_score: 0.5
      threshold: 0.7
    bus:
      min_score: 0.5
      threshold: 0.7
    bird:
      min_score: 0.5
      threshold: 0.7
    cat:
      min_score: 0.65
      threshold: 0.8
    dog:
      min_score: 0.7
      threshold: 0.9
    hat:
      min_score: 0.5
      threshold: 0.7
    backpack:
      min_score: 0.5
      threshold: 0.7
    umbrella:
      min_score: 0.5
      threshold: 0.7
    eye glasses:
      min_score: 0.5
      threshold: 0.7
    handbag:
      min_score: 0.5
      threshold: 0.7
    suitcase:
      min_score: 0.5
      threshold: 0.7
    face:
      min_score: 0.7
      threshold: 0.7
    license_plate:
      min_score: 0.6
      threshold: 0.7
    amazon:
      min_score: 0.75
      threshold: 0.7
    fedex:
      min_score: 0.75
      threshold: 0.7
    ups:
      min_score: 0.75
      threshold: 0.7
    package:
      min_score: 0.65
      threshold: 0.9
    #dhl:
    #  min_score: 0.65
    #  threshold: 0.9
    #dpd:
    #  min_score: 0.65
    #  threshold: 0.9
    #fox:
    #  min_score: 0.65
    #  threshold: 0.9
    #squirrel:
    #  min_score: 0.65
    #  threshold: 0.9
    #waste_bin:
    #  min_score: 0.65
    #  threshold: 0.9

record:
  #sync_recordings: true # = synchronises files which may have been deleted outside of Frigate; uses considerable CPU resources and is mostly not needed.
  enabled: true
  retain:
    days: 14
    mode: motion
  events:
    retain:
      default: 14
      mode: all #- all records all segments of an event start to finish, motion=only motion during event, active_objects=only when objects active (car/people moving etc)
      #objects: # - retention policy for individual objects
      #  dog: 7
      #  person: 14
      #  package: 10
  export:
    timelapse_args: -vf setpts=0.02*PTS -r 25

#review:
  # global objects to produce alerts
  #alerts:
    #labels:
      #- person
      #- car
      #- bicycle
      #- motorcycle
      #- bus
      #- bird
      #- cat
      #- dog
      #- horse
      #- bear
      #- hat
      #- backpack
      #- umbrella
      #- eye glasses
      #- handbag
      #- suitcase
      #- face
      #- license_plate
      #- amazon
      #- ups
      #- fedex
      #- package
      #- dhl
      #- dpd
      #- fox
      #- squirrel
      #- waste_bin

snapshots:
  enabled: true
  clean_copy: true
  timestamp: false
  bounding_box: true
  retain:
    default: 14
    objects:
      person: 30
      car: 14
      bicycle: 30
      bus: 14
      bird: 30
      cat: 30
      dog: 30

cameras:
  FrontPath:
    detect:
      #width: 704
      #height: 576
      fps: 10
    enabled: true
    ffmpeg:
      inputs:
        #- path: rtsp://127.0.0.1:8554/FrontPath_Sub
        #  input_args: preset-rtsp-restream
          #roles:
          #  - detect
        - path: rtsp://127.0.0.1:8554/FrontPath_Sub
          input_args: preset-rtsp-restream
          roles:
            - detect
        - path: rtsp://127.0.0.1:8554/FrontPath
          input_args: preset-rtsp-restream
          roles:
            - record
    objects:
      filters:
        person:
          #min_area: 2000
          #max_area: 14000
          #min_ratio: 0.25
          #max_ratio: 0.8
          min_score: 0.72
          threshold: 0.7
          #min_score: 0.15
          #threshold: 0.1
    zones:
      FrontPathway:
        coordinates: 0.604,0.502,0.793,0.478,0.937,0.541,0.794,0.999,0.425,0.997
        loitering_time: 0
        inertia: 3
        #filters:
        #  person:
        #    min_area: 2000
        #    max_area: 14000
        #    min_score: 0.5
        #    threshold: 0.52
        objects:
          - backpack
          - bicycle
          - bird
          - cat
          - dog
          - eye glasses
          - handbag
          - hat
          - package
          - person
          - suitcase
          - umbrella
      CarTrack:
        coordinates: 0.763,0.175,0.975,0.328,0.95,0.498,0.627,0.305
        loitering_time: 0
        objects:
          - car
          - motorcycle
        inertia: 3
    review:
      alerts:
        labels:
          - backpack
          - bicycle
          - bird
          - cat
          - dog
          - eye glasses
          - handbag
          - hat
          - package
          - person
          - suitcase
          - umbrella
        #required_zones: FrontPathway
    motion:
      threshold: 30
      contour_area: 10
      improve_contrast: 'true'
      mask:
        - 0.195,0.914,0.213,0.911,0.212,0.945,0.195,0.944
        - 0.465,0.003,0.45,0.042,0.424,0.109,0.404,0.176,0.389,0.2,0.385,0.273,0.297,0.407,0.359,0.511,0.373,0.595,0.323,0.642,0.321,0.709,0.362,0.764,0.4,0.788,0.453,0.836,0.512,0.812,0.573,0.834,0.586,0.757,0.589,0.7,0.583,0.649,0.583,0.605,0.59,0.563,0.593,0.531,0.601,0.496,0.599,0.452,0.593,0.423,0.598,0.393,0.595,0.369,0.601,0.324,0.599,0.299,0.605,0.281,0.568,0.268,0.568,0.241,0.599,0.188,0.562,0.197,0.556,0.161,0.552,0.136,0.566,0.103,0.529,0.076,0.52,0.072,0.52,0.041,0.51,0.026,0.496,0.014,0.48,0.012
        - 0.008,0.636,0.043,0.651,0.07,0.994,0.006,0.997

version: 0.14

docker-compose file or Docker CLI command

N/A

Relevant Frigate log output

Trying qsv gives me errors like: 
`Ffmpeg process crashed unexpectedly for FrontPath.
2024-11-23 18:16:07.540300350  [2024-11-23 18:16:07] watchdog.FrontPath             ERROR   : The following ffmpeg logs include the last 100 lines prior to exit.
2024-11-23 18:16:07.540360396  [2024-11-23 18:16:07] ffmpeg.FrontPath.detect        ERROR   : [AVHWDeviceContext @ 0x55d77a47b680] No VA display found for device -hwaccel_output_format.
2024-11-23 18:16:07.540421443  [2024-11-23 18:16:07] ffmpeg.FrontPath.detect        ERROR   : Device creation failed: -22.
2024-11-23 18:16:07.540460040  [2024-11-23 18:16:07] ffmpeg.FrontPath.detect        ERROR   : Failed to set value '-hwaccel_output_format' for option 'qsv_device': Invalid argument
2024-11-23 18:16:07.540500334  [2024-11-23 18:16:07] ffmpeg.FrontPath.detect        ERROR   : Error parsing global options: Invalid argument`

Using environment vars i965 and vaapi gives:
`[2024-11-23 19:25:39] frigate.util.services          ERROR   : Unable to poll intel GPU stats: No device filter specified and no discrete/integrated i915 devices found
2024-11-23 19:25:39.325736919  
2024-11-23 19:25:39.604099052  [2024-11-23 19:25:39] frigate.video                  ERROR   : FrontPath: Unable to read frames from ffmpeg process.`
and 
`[2024-11-23 19:25:58] ffmpeg.FrontPath.detect ERROR   : Option hwaccel (use HW accelerated decoding) cannot be applied to output url vaapi -- you are trying to apply an input option to an output file or vice versa. Move this option before the file it belongs to.
2024-11-23 19:25:58.894778812  [2024-11-23 19:25:58] watchdog.FrontPath             ERROR   : The following ffmpeg logs include the last 100 lines prior to exit.
2024-11-23 19:25:58.894966237  [2024-11-23 19:25:58] ffmpeg.FrontPath.detect ERROR   : Error parsing options for output file vaapi.
2024-11-23 19:25:58.894972777  [2024-11-23 19:25:58] ffmpeg.FrontPath.detect ERROR   : Error opening output files: Invalid argument
2024-11-23 19:25:58.894979279  [2024-11-23 19:25:58] ffmpeg.FrontPath.detect        ERROR   : Option hwaccel (use HW accelerated decoding) cannot be applied to output url vaapi -- you are trying to apply an input option to an output file or vice versa. Move this option before the file it belongs to.`

Just using vaapi:
`[2024-11-23 19:30:28] frigate.video                  ERROR   : FrontPath: Unable to read frames from ffmpeg process.
2024-11-23 19:30:28.648519630  [2024-11-23 19:30:28] frigate.video                  ERROR   : FrontPath: ffmpeg process is not running. exiting capture thread...
[2024-11-23 19:30:29] frigate.util.services          ERROR   : Unable to poll vainfo: b"error: XDG_RUNTIME_DIR is invalid or not set in the environment.\nerror: can't connect to X server!\nerror: failed to initialize display\n"`

Relevant go2rtc log output

N/A

FFprobe output from your camera

don't know how to do this, sorry

Install method

HassOS Addon

Object Detector

Coral

Network connection

Wired

Camera make and model

milesight/dahua

Screenshots of the Frigate UI's System metrics pages

Any other information that may be helpful

No response

[iamhermes]

Is there a connected screen/display? If not, please add a screen and reboot the os (if kernel boot parameter "nomodeset" is present, please remove it)
Linux will disable the GPU if there is no screen attached.

[matknowles]

Thanks very much for helping me out. There currently isn’t a screen connected, I’ll sort that ASAP. I don’t usually use a screen so is this something I’d need to keep connected constantly or can I just keep an HDMI cable connected to a display that’s in use for other things etc? Can you also explain how to check the kernel boot parameters please? I’m running home assistant os generic x86-64 14.0.rc2 directly from the nvme if it matters. Thanks, Matthew From: iamhermes ***@***.***> Sent: 23 November 2024 21:17 To: blakeblackshear/frigate ***@***.***> Cc: matknowles ***@***.***>; Author ***@***.***> Subject: Re: [blakeblackshear/frigate] [HW Accel Support]: Asus NUC 14 Pro / Intel Core Ultra 7 155H (Discussion #15153) Is there a connected screen/display? If not, please add a screen and reboot the os (if kernel boot parameter "nomodeset" is present, please remove it) Linux will disable the GPU if there is no screen attached. — Reply to this email directly, view it on GitHub<#15153 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/BESDYQ5MQDIJVHO2T5SVPST2CDWD7AVCNFSM6AAAAABSLMSXUOVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTCMZVHE3DSMY>. You are receiving this because you authored the thread.Message ID: ***@***.***>

[iamhermes]

Look for an HDMI dummy plug. These simulate a monitor and thereby activate the GPU.
There is no need for a real monitor to be connected. And if there is a real monitor, it does not necessarily have to be connected to the power supply.

[matknowles]

OK I've inserted a dummy plug, rebooted the system, but am still getting the same errors. I've tried frigate 0.15.2 beta, but am also getting errors with OPENVINO.

WARNING : OpenVINO failed to build model, using CPU instead: Exception from src/inference/src/cpp/core.cpp:121: 2024-11-24 10:43:21.634758811 Check '!m_device_map.empty()' failed at src/plugins/intel_gpu/src/plugin/plugin.cpp:379:
I'll attach a logfile incase there's anything else I'm not seeing.
frigate beta log with environment vars i965 and vaapi.log
frigate beta log with environment vars i965 and qsv.log

The dummy plug I'm using is "DTECH HDMI Dummy Plug 4K Display Emulator Compatible with Windows Mac OSX Linux (fit-Headless, 3 pack)" from Amazon

[matknowles]

couple more logs with environment vars disabled and vaapi/qsv if it helps.
frigate beta logs no environment vars, qsv enabled.log
frigate beta log no environment vars, vaapi enabled.log

[iamhermes]

I don't know your host os. But will this answer help further?
https://community.home-assistant.io/t/problem-with-haos-and-frigate-hwaccel/513290

[jake9190]

What's your underlying OS and are you running in something like HAOS? We need to know what your passthrough configuration looks like for the GPU

I have a NUC 14 Pro but with the i3-100u. I had similar difficulties, and there are a few posts about getting this to work on proxmox in some other places. If I remember correctly, Intel dropped IOMMU support in later CPUs, and only supports SR-IOV for passthrough, but support is pretty poor in most distros right now. I was eventually able to get it to work on Debian, but I'm not 100% sure I can reproduce it.

See if your GPU shows up under /dev/dri first, if it does, the problem is probably somewhere else. I see card0 and renderD128. If you see those, you don't have the same issue that I did. To get this to work on Debian, I had to:

1. Install an updated kernel (I'm using xanmod 6.11)
2. Install linux-image-amd64 linux-headers-amd64 from backports
3. Install linux-firmware
4. Installing intel-media-va-driver-non-free/unstable (I would probably just build this if I had to do it again)
5. I may have also made a change to get card0 to show up. I think it was a boot param.

This enabled "sudo intel_gpu_top -d drm:/dev/dri/renderD128" to work, and passthrough to my docker container.

At some point I may buy another NUC so I can play around and see if I can repro getting it to work.

[matknowles]

Thanks for your help. I'm running home assistant Operating System 14.0.rc2 (used the linux mint Disks restore disk image install method - haos_generic-x86-64-14.0.rc2.img.xz and I boot directly to that, Core 2024.11.3, Supervisor 2024.11.4, Frontend 20241106.2. Frigate beta 0.15.0b2.

When i unboxed the NUC and tried to boot it, it froze immediately and wouldn't boot any linux images. I googled the problem and found I had to disable intel VT for directed I/O (VT-d), which worked. I've looked for other relevant BIOS settings but don't see any. As suggested I've tried a HDMI dummy plug to try and activate the GPU on boot but it doesn't seem to have worked.

I've tried to add ffmpeg: to my configuration.yaml, but am not experienced enough to diagnose the version/gpu/boot parameters etc, someone advised to disable nomodeset but I don't know how to do that.

I'm considering installing debian 12.8 on my NUC then running HAOS/frigate etc in there, but am inexperienced with docker and linux in general, so would rather do that as a last resort. Also I might try the stable version of HAOS 13.2 to see if that fixes anything.

"See if your GPU shows up under /dev/dri first", can I do this in HAOS? I have the terminal addon but am not sure how it works. I've also tried diagnosing ffmpeg issues in the terminal addon but am getting "ffmpeg: command not found"

Thanks again for your help, really appreciate it. I hope I don't have to return this NAS to get an older one that works better, but am happy to keep working to try and fix it. As I mentioned, it seems relatively happy without hardware acceleration at the moment; roughly 18-26% CPU usage with semantic search etc enabled, but I know it's not ideal.

[jake9190]

OK then your issue is different than mine since you're not running in a VM or container (I'm not familiar with HAOS or mint, so I might be wrong with how HAOS is setup).

A couple of things to try:

1. At the HAOS terminal, rung: 'ls /dev/dri' and see what the results are. You should see the two entries I mentioned above.
2. Install and run either vainfo, or intel_gpu_top. For mint this should be: 'sudo apt install vainfo intel-gpu-tools'

These will tell you if the GPU is accessible to the OS, but won't fix the issue.

I do wonder if disabling VT-d is related. I've founded Asus support to be decent, so it may be worth reaching out to them to see why you need to disable VT-d to boot.

[matknowles]

Apologies, I didn't explain properly. I only used mint as a live boot initially once to "expand" the home assistant os image to my NUCs nvme (using the Disks app), as recommended on home assistant's page for installing on generic x86-64 hardware.

As an update I tried your commands in home assistant terminal and got no such file or directory for ls /dev/dri and command not found: vainfo. I tried booting to debian 12.8 live by USB; it didn't like it, throwing some errors about not being able to load wifi/firmware and bluetooth but got to a desktop in what looked like 640x480, ran the commands in there and still got no such file or directory for ls /dev/dri (I wonder if the OS needs to be installed for this to work? *ignore this/read on)) and vainfo gave errors (will edit/update with these I've forgotten them -.- - just saw you said to install it, I didn't do that. again ignore this bit) I'm not sure I can install it on HAOS; apt command isn't found either and I've heard HAOS is pretty locked down for various reasons.)

I also messed around in BIOS and re-enabled VT-d and some other settings; I have no idea whether any are related:
Enabled internal UEFI shell
Enabled Extend CSME measurement to TPM-PCR
Increased IGD Minimum memory to 128MB
Enabled Allow UEFI 3rd party driver to be loaded (could well be this one that's fixed the boot issue with VT-d?)
Enabled Intel Virtualisation technology (if this wasn't enabled already; can't remember)
Enabled Intel VT for Directed I/O (VT-d)
Set iSetupCfg password check to Bypass

and managed to get the systems to boot properly, so I have no idea what's happened there, but unfortunately I still can't get the hwaccel working in frigate beta in HAOS with any combination of environment_vars and hwaccel_args, it was the first thing I excitedly checked, then was immediately deflated :/

Just noticed you asked to install/run intel_gpu top... again it's not found on HAOS.

I may try and roll back to haos stable 13.2 instead of 14.0.rc2 on the offchance that fixes something.
I've also reached out to ASUS for their help with it.
As a last resort I might try installing debian to the nvme and running frigate in docker, but seeing how bumpy the live boot was, I'm not confident in that as a solution.will probably choose mint if anything considering the below experiment

I booted to linux mint as a live usb and installed/ran vainfo and ran ls /dev/dri and got the following:
vainfo libva info: VA-API version 1.20.0 libva info: Trying to open /usr/lib/x86_64-linux-gnu/dri/iHD_drv_video.so libva info: Found init function __vaDriverInit_1_20 libva info: va_openDriver() returns 0 vainfo: VA-API version: 1.20 (libva 2.12.0) vainfo: Driver version: Intel iHD driver for Intel(R) Gen Graphics - 24.1.0 () vainfo: Supported profile and entrypoints VAProfileNone : VAEntrypointVideoProc VAProfileMPEG2Simple : VAEntrypointVLD VAProfileMPEG2Main : VAEntrypointVLD VAProfileH264Main : VAEntrypointVLD VAProfileH264Main : VAEntrypointEncSlice VAProfileH264High : VAEntrypointVLD VAProfileH264High : VAEntrypointEncSlice VAProfileJPEGBaseline : VAEntrypointVLD VAProfileJPEGBaseline : VAEntrypointEncPicture VAProfileH264ConstrainedBaseline: VAEntrypointVLD VAProfileH264ConstrainedBaseline: VAEntrypointEncSlice VAProfileVP8Version0_3 : VAEntrypointVLD VAProfileHEVCMain : VAEntrypointVLD VAProfileHEVCMain : VAEntrypointEncSlice VAProfileHEVCMain10 : VAEntrypointVLD VAProfileHEVCMain10 : VAEntrypointEncSlice VAProfileVP9Profile0 : VAEntrypointVLD VAProfileVP9Profile0 : VAEntrypointEncSlice VAProfileVP9Profile1 : VAEntrypointVLD VAProfileVP9Profile1 : VAEntrypointEncSlice VAProfileVP9Profile2 : VAEntrypointVLD VAProfileVP9Profile2 : VAEntrypointEncSlice VAProfileVP9Profile3 : VAEntrypointVLD VAProfileVP9Profile3 : VAEntrypointEncSlice VAProfileHEVCMain12 : VAEntrypointVLD VAProfileHEVCMain422_10 : VAEntrypointVLD VAProfileHEVCMain422_12 : VAEntrypointVLD VAProfileHEVCMain444 : VAEntrypointVLD VAProfileHEVCMain444 : VAEntrypointEncSlice VAProfileHEVCMain444_10 : VAEntrypointVLD VAProfileHEVCMain444_10 : VAEntrypointEncSlice VAProfileHEVCMain444_12 : VAEntrypointVLD VAProfileHEVCSccMain : VAEntrypointVLD VAProfileHEVCSccMain : VAEntrypointEncSlice VAProfileHEVCSccMain10 : VAEntrypointVLD VAProfileHEVCSccMain10 : VAEntrypointEncSlice VAProfileHEVCSccMain444 : VAEntrypointVLD VAProfileHEVCSccMain444 : VAEntrypointEncSlice VAProfileAV1Profile0 : VAEntrypointVLD VAProfileAV1Profile0 : VAEntrypointEncSlice VAProfileHEVCSccMain444_10 : VAEntrypointVLD VAProfileHEVCSccMain444_10 : VAEntrypointEncSlice mint@mint:~$ ls /dev/dri by-path card1 renderD128

As it happens, everything seemed to be working in mint live as well; bluetooth,wifi, full resolution graphics, so it's perhaps either kernel or intel firmware related? I'm not knowledgable enough though.

[matknowles]

OK thanks for your advice. Does the distro matter? From my live boot experiments, Debian gave lots of errors relating to intel firmware/bluetooth/wifi and had a low display resolution. Mint seemed to have everything working straight away, so would it be a better starting point? Apologies for going off topic.

[NickM-27]

It doesn't strictly matter as far as Frigate is concerned but our ability to help you may be affected some, though mint is based on debian so shouldn't be an issue

[iamhermes]

Is the resolution important on a headless server? All maintenance is done via the web or SSH.
In my use case, I explicitly looked for Intel AMT-enabled devices so that I could perform a reset if necessary or see the last screen output without accessing the device.
However, a dummy plug or connected screen is also necessary here because of Linux, otherwise there will be no screen output. In addition to the issue that the GPU does not activate if there is no screen recognition. This is not an issue when using Microsoft Windows as the main operating system.
However, this now only applies to the remote KVM.

Thank you for the tip about dockge, I am currently using portainer and find it overloaded with too few responses for my purposes. I will soon switch back to my first hardware after my freeze problem (most likely defective dummy plug) has now been solved and will use dockge here straight away.

I also had HomeAssistant OS running. This ran on an Ubuntu Server HostOS, on which I installed cockpit to create/control the VM for home assistant OS. As a VM to be able to use the full range of functions of Home Assistant.I don't know if there are any easier/better ways to set up a VM for HA OS.

[matknowles]

of course it doesn't matter per-se, but it may be indicative of a missing driver or incompatible firmware, especially as far as hardware acceleration is concerned, which pertains to the issue I'm having. My thinking is by using a distro where those things are already working, it's less likely I'm going to experience issues related to them.

Home assistant's site reccomends to use virtualbox for installing HAOS within linux, but it's not something I have experience with or knowledge of, so I'll have to learn as I go, hopefully it's not too bad. I'll try to get to it tonight.

[iamhermes]

Ubuntu server as bare metal, run frigate through dockge (web management), HA OS via kimchi or similar (is it good enough?)
Or take proxmox (Debian) or another distro.
Ubuntu or Debian, it's easier to find help if there are problems.

HA OS doesn't include much packages or drivers, because of support.