[HW Accel Support]: Intel Arc380 gpu accel working with ov detector but not onnx #20358
Unanswered
supraway
asked this question in
Hardware Acceleration Support
Replies: 1 comment 10 replies
-
This is generally known issue as there is some bug with the onnx runtime. Recommendation for detector is to use openvino, that's what the enrichments do behind the scenes |
Beta Was this translation helpful? Give feedback.
10 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
Describe the problem you are having
I've been having trouble getting hardware acceleration working using the onnx detector with my Intel Arc380 GPU on a system running docker compose.
Previously, Frigate has been working well with verified hardware acceleration (both ffmpeg and detector) and sub- 10ms inference times using YOLO-NAS with openvino. I wanted to give RF-DETR a try (compiled without error on the same machine as Frigate and per wiki instructions) so I updated my config and switched detectors to onnx, and I'm no longer seeing a detector worker in intel_gpu_top, high cpu usage, and 60+ms inference times.
I've tried several combinations detector/model to test things:
I also tried all these detector/model combos running frigate in privileged and unprivileged mode in docker compose, and tried preset-vaapi and preset-intel-qsv-h264 without change.
I included the logs for the openvino error (configuration #3 above) hoping maybe that will give some clues. There are no errors with the other configurations, and does confirm "ONNX: /config/model_cache/rfdetr-Nano.onnx loaded".
Version
0.16.1-e664cb2
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
Docker Compose
Object Detector
Other
Network connection
Wired
Camera make and model
amcrest
Screenshots of the Frigate UI's System metrics pages
Any other information that may be helpful
intel_gpu_top output:
Kernel info:
Linux xxxxxx 6.12.43+deb12-amd64 #1 SMP PREEMPT_DYNAMIC Debian 6.12.43-1~bpo12+1 (2025-09-06) x86_64 GNU/Linux
Beta Was this translation helpful? Give feedback.
All reactions