Add Alder Lake configuration for hwaccel

This commit is contained in:
Charles Munger 2022-07-25 15:33:09 -07:00
parent 37325c70ba
commit 9736a23804

View File

@ -15,7 +15,11 @@ ffmpeg:
hwaccel_args: -c:v h264_v4l2m2m
```
### Intel-based CPUs (<10th Generation) via Quicksync
### Intel-based CPUs with QuickSync
Intel CPUs with build in graphics (non-F variants) since 5th generation have at least h264 decoding support. Once you've configured it and restarted frigate, verify that it's working by running `intel_gpu_top` - if everything is set up correctly, you should see non-zero load on `Video` and/or `VideoEnhance`.
#### <10th Generation via `vaapi`
```yaml
ffmpeg:
@ -23,13 +27,38 @@ ffmpeg:
```
**NOTICE**: With some of the processors, like the J4125, the default driver `iHD` doesn't seem to work correctly for hardware acceleration. You may need to change the driver to `i965` by adding the following environment variable `LIBVA_DRIVER_NAME_JELLYFIN=i965` to your docker-compose file.
### Intel-based CPUs (>=10th Generation) via Quicksync
### >=10th Generation via Quicksync
```yaml
ffmpeg:
hwaccel_args: -c:v h264_qsv
```
### >=12th Generation via Quicksync
**Note:** You should use kernel 5.18 or newer for full support.
```yaml
ffmpeg:
input_args:
- -c:v
- h264_qsv # Or h265_qsv depending on your camera's format
- -gpu_copy
- "on"
...
hwaccel_args:
- -hwaccel
- qsv
- -qsv_device
- /dev/dri/renderD128
- -hwaccel_output_format
- qsv
output_args:
# Converts from the format used on the GPU to the format expected by frigate.
detect: -vf vpp_qsv=format=yuv420p -f rawvideo -pix_fmt yuv420p
```
Specifying `input_args` and `output_args` overrides frigate's defaults, so you may want to append whatever options your camera requires.
### AMD/ATI GPUs (Radeon HD 2000 and newer GPUs) via libva-mesa-driver
**Note:** You also need to set `LIBVA_DRIVER_NAME=radeonsi` as an environment variable on the container.