GPU usage is always zero #16120
-
Describe the problem you are havingI'm using Frigate for some time now. I noticed in the stats page that my GPU always indicates no usage (0.0%). I'm running Frigate on a HardKernel Odroid H4 Ultra, Openmediavault 7.5.1-1 compose Version14.1 Frigate config fileui:
time_format: 24hour
strftime_fmt: '%d/%m/%Y %H:%M:%S'
mqtt:
enabled: false
detectors:
coral:
type: edgetpu
device: usb
database:
path: /media/frigate/frigate.db
#logger:
# default: info
# logs:
# frigate.record: debug
birdseye:
enabled: false
mode: continuous
ffmpeg:
hwaccel_args: preset-vaapi
input_args: preset-rtsp-restream
output_args:
record: preset-record-generic-audio-aac
record:
enabled: true
sync_recordings: false
retain:
days: 10
mode: all
events:
retain:
default: 10
go2rtc:
streams:
tuinhuiscam:
- rtsp://rtspuser:[email protected]:554/Streaming/Channels/101
tuinhuiscam_sub:
- rtsp://rtspuser:[email protected]:554/Streaming/Channels/102
deurbel:
- ffmpeg:http://192.168.1.65/flv?port=1935&app=bcs&stream=channel0_main.bcs&user=admin&password=xxxxxxxx#video=copy#audio=copy#audio=opus
deurbel_sub:
- ffmpeg:http://192.168.1.65/flv?port=1935&app=bcs&stream=channel0_ext.bcs&user=admin&password=xxxxxxxx
cameras:
tuinhuiscam: # geen output_args: geen audio
enabled: true
ffmpeg:
inputs:
- path: rtsp://localhost:8554/tuinhuiscam?video&audio
roles:
- record
- path: rtsp://localhost:8554/tuinhuiscam_sub
roles:
- detect
detect:
enabled: true
objects:
track:
- person
motion:
mask:
- 0,0.321,0.308,0.195,0.307,0,0,0
- 1,0,1,1,0.962,1,0.858,0
- 0.815,0.441,0.849,0.442,0.849,0.152,0.815,0.151
- 0,0.323,0.077,0.291,0.082,1,0,1
zones:
tuin:
coordinates: 0.961,1,0.873,0.164,0.388,0.166,0,0.324,0,1
objects: person
inertia: 3
loitering_time: 0
review:
alerts:
required_zones: tuin
deurbel: # <--- this will be changed to your actual camera later
enabled: true
ffmpeg:
inputs:
- path: rtsp://localhost:8554/deurbel
roles:
- record
- path: rtsp://localhost:8554/deurbel_sub
roles:
- detect
detect:
enabled: true
objects:
track:
- person
- bicycle
- motorcycle
motion:
mask: 0,0,1,0,1,0.413,0,0.413
zones:
straat:
coordinates: 0,1,1,1,1,0.414,0,0.414
objects:
- person
- bicycle
- motorcycle
inertia: 3
loitering_time: 0
review:
alerts:
required_zones: straat
version: 0.14
camera_groups: {}Relevant Frigate log output2025-01-23 19:44:41.614384595 [INFO] Preparing Frigate...
2025-01-23 19:44:41.629946464 [INFO] Starting Frigate...
2025-01-23 19:44:43.023575141 [2025-01-23 19:44:43] frigate.app INFO : Starting Frigate (0.14.1-f4f3cfa)
2025-01-23 19:44:43.023608418 [2025-01-23 19:44:43] frigate.util.config INFO : Checking if frigate config needs migration...
2025-01-23 19:44:43.045166453 [2025-01-23 19:44:43] frigate.util.config INFO : frigate config does not need migration...
2025-01-23 19:44:55.405576938 [2025-01-23 19:44:55] peewee_migrate.logs INFO : Starting migrations
2025-01-23 19:44:55.405713468 [2025-01-23 19:44:55] peewee_migrate.logs INFO : There is nothing to migrate
2025-01-23 19:44:55.426717633 [2025-01-23 19:44:55] frigate.app INFO : Recording process started: 389
2025-01-23 19:44:55.432271876 [2025-01-23 19:44:55] frigate.app INFO : Recording process started: 395
2025-01-23 19:44:55.436011066 [2025-01-23 19:44:55] frigate.app INFO : go2rtc process pid: 98
2025-01-23 19:44:55.455842451 [2025-01-23 19:44:55] detector.coral INFO : Starting detection process: 417
2025-01-23 19:44:55.458825736 [2025-01-23 19:44:55] frigate.app INFO : Output process started: 419
2025-01-23 19:44:55.510036070 [2025-01-23 19:44:55] frigate.app INFO : Camera processor started for tuinhuiscam: 436
2025-01-23 19:44:55.510121679 [2025-01-23 19:44:55] frigate.app INFO : Camera processor started for deurbel: 439
2025-01-23 19:44:55.517653414 [2025-01-23 19:44:55] frigate.app INFO : Capture process started for tuinhuiscam: 443
2025-01-23 19:44:55.519100171 [2025-01-23 19:44:55] frigate.app INFO : Capture process started for deurbel: 453
2025-01-23 19:44:58.099822138 [2025-01-23 19:44:55] frigate.detectors.plugins.edgetpu_tfl INFO : Attempting to load TPU as usb
2025-01-23 19:44:58.105723692 [2025-01-23 19:44:58] frigate.detectors.plugins.edgetpu_tfl INFO : TPU foundRelevant go2rtc log output2025-01-23 19:44:41.621072254 [INFO] Preparing new go2rtc config...
2025-01-23 19:44:42.048059803 [INFO] Starting go2rtc...
2025-01-23 19:44:42.200435034 19:44:42.200 INF go2rtc platform=linux/amd64 revision=b2399f3 version=1.9.2
2025-01-23 19:44:42.200439737 19:44:42.200 INF config path=/dev/shm/go2rtc.yaml
2025-01-23 19:44:42.200985150 19:44:42.200 INF [api] listen addr=:1984
2025-01-23 19:44:42.201328330 19:44:42.201 INF [rtsp] listen addr=:8554
2025-01-23 19:44:42.201529911 19:44:42.201 INF [webrtc] listen addr=:8555/tcp
2025-01-23 19:44:51.614775659 [INFO] Starting go2rtc healthcheck service...Frigate stats---Operating systemOther Linux Install methodDocker Compose docker-compose file or Docker CLI command---
# https://github.com/blakeblackshear/frigate
services:
frigate:
container_name: frigate
privileged: true # this may not be necessary for all setups
restart: unless-stopped
image: ghcr.io/blakeblackshear/frigate:stable
# image: ghcr.io/blakeblackshear/frigate:0.13.2
shm_size: "160mb" # 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/dri/renderD128 # for intel hwaccel, needs to be updated for your hardware
volumes:
- /etc/localtime:/etc/localtime:ro
- ${PATH_TO_APPDATA}/frigate/config.yml:/config/config.yml
- ${PATH_TO_MOVIES}/frigate/storage:/media/frigate
- type: tmpfs # Optional: 1GB of memory, reduces SSD/SD Card wear
target: /tmp/cache
tmpfs:
size: 1000000000
ports:
- "5000:5000"
- "8554:8554" # RTSP feeds
- "8555:8555/tcp" # WebRTC over tcp
- "8555:8555/udp" # WebRTC over udp
environment:
FRIGATE_RTSP_PASSWORD: "password"Object DetectorCoral Screenshots of the Frigate UI's System metrics pages
Any other information that may be helpfulNo response |
Beta Was this translation helpful? Give feedback.
Replies: 12 comments 15 replies
-
|
in your case the coral is used for detection, not the GPU. you can run |
Beta Was this translation helpful? Give feedback.
-
|
I ran
|
Beta Was this translation helpful? Give feedback.
-
|
right, like I said you should run it on the host. would also be good to see hardware info from the metrics screen |
Beta Was this translation helpful? Give feedback.
-
|
You mean this?
|
Beta Was this translation helpful? Give feedback.
-
|
looks like the driver is correct, so it should be using the GPU. If you run Frigate 0.15 then you will want to use qsv and that may show up on the stats |
Beta Was this translation helpful? Give feedback.
-
|
0.15 is not yet official.
into this
Because I experimented with this in 0.14 and it wasn't a success (no video from cameras) |
Beta Was this translation helpful? Give feedback.
-
|
yes that is the only change that is needed |
Beta Was this translation helpful? Give feedback.
-
|
Just to roundup this issue.
|
Beta Was this translation helpful? Give feedback.
-
|
Did it stay this way? I'm seeing similar stuff on my N150 machine. GPU is used... stats are correct and after a night it drops to 0%.
after machine reboot GPU usage is once again reported correctly in frigate |
Beta Was this translation helpful? Give feedback.
-
|
After a reboot of my NAS (i3-N305; OMV latest, Frigate 0.15 compose) all seems fine. In
But after some time
Waiting some more time and the activity of
Remark: The typical Frigate processes ffmpeg, go2rtc, nginx and frigate.recording/process/output/capture/review under Until now I found:
Looking at your setup it seems to rule out that the cause is the hardware or the host OS. Maybe still an Intel driver issue. |
Beta Was this translation helpful? Give feedback.
-
Can reproduce the same exact behavior on A380 inside Unraid, where Frigate reports some usage and then silence.... |
Beta Was this translation helpful? Give feedback.
-
|
Same here on my Arc a310 |
Beta Was this translation helpful? Give feedback.
-
@Imwithstupid007 rest is the same as You described |
Beta Was this translation helpful? Give feedback.
-
|
having a similar issue. I am trying to utilize two GPUs, an intel 12th gen iGPU and an arc a310 within unraid. Frigate shows constant 0% with GPU, but I do see GPU usage in intel_gpu_top. Is hardware accel and openvino working all good and this is just a UI issue?
|
Beta Was this translation helpful? Give feedback.
-
|
I also saw this: My container is running privileged and s |
Beta Was this translation helpful? Give feedback.
-
|
This is a bug with Intel GPU tools. Nothing we can work around at this point. |
Beta Was this translation helpful? Give feedback.
-
Thanks for that info. Wondering if it's possible to have a status page with information if hardware acceleration and GPU detection is working correctly. I spent all day trying to figure out why the GPU usage wasn't showing anything in frigate |
Beta Was this translation helpful? Give feedback.
-
|
This is what the GPU usage is intended to show. There's no other way to directly show that it's working, but if your inference speeds and CPU usage remains low and there are no errors in the logs, you can safely assume it's still working even after it drops to zero. This likely will be fixed in a later version of Frigate when we can safely update all of the appropriate Intel packages. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks, understood. I think my CPU usage is a bit higher because I am still using 5 nest cams within frigate (transitioning to poe cams, slowly) and they have no lower res sub stearm |
Beta Was this translation helpful? Give feedback.
-
|
I think I fixed it, at least on the arc a310 - I booted up wintoUSB on the server with the arc GPU installed, installed the latest arc GPU drivers. When do, it tells you that it will also update the firmware. After updating, booted back into unraid and now frigate and unraid correctly shows GPU usage
not sure if it works for intel iGPUs |
Beta Was this translation helpful? Give feedback.
-
|
Is it showing for longer than an hour? |
Beta Was this translation helpful? Give feedback.
-
nope, meant to update - seems that it was similar behavior to the others: after reboot, it works for a bit, then it goes back to reporting 0% all the time |
Beta Was this translation helpful? Give feedback.
-
|
Same issue here—Frigate 14.1 in a container on a Beelink S12, and my iGPU just sits at 0% usage. A reboot gets things moving again, but after a few hours, it drops back to zero like clockwork. Hopefully this is just a visual glitch and not a hardware problem. |
Beta Was this translation helpful? Give feedback.
-
|
Same issue shows 0% if i restart frigate still shows 0% are you guys rebooting the host to get it to shows initially? |
Beta Was this translation helpful? Give feedback.
-
yep. Host restart required. |
Beta Was this translation helpful? Give feedback.
-
|
Hi, IMHO this bug shoud not be closed as the behavior is still existing with latest stable frigate version at date : Intel GPU show some usage for some time and then get to 0% in graphs under System Stats. Configuration does not change, Host reboot solved the probleme but only for a few hours and then again stopped. What to do with this behavior please ? Regards, Laurent |
Beta Was this translation helpful? Give feedback.
-
|
This is not a Frigate issue, there is nothing Frigate itself can do about this |
Beta Was this translation helpful? Give feedback.
Just to roundup this issue.
After a reboot of the NAS things look like they should be.