administrator/frigate
1
0
Fork 0
mirror of https://github.com/blakeblackshear/frigate.git synced 2026-10-19 21:22:50 +03:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

Upgrade to ROCm 10.1 (#24708)

Browse Source 
* Update to ROCm 10.1

* Update docs for new setup
This commit is contained in:
Nicolas Mowen
2026-10-18 12:13:32 -05:00
committed by GitHub
parent 893542f4ca
commit d7bb780ff8
9 changed files with 410 additions and 103 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/docs/configuration/object_detectors.md
+26 -18
View File
@@ -394,6 +394,22 @@ Note that the labelmap uses a subset of the complete COCO label set that has onl
Support for AMD GPUs is provided using the [ONNX detector](#onnx). In order to utilize the AMD GPU for object detection use a frigate docker image with `-rocm` suffix, for example `ghcr.io/blakeblackshear/frigate:stable-rocm`.
### Supported GPUs {#supported-gpus-rocm}
The `-rocm` image includes ROCm support for these GPU chipsets, which work without any chipset override:
| Chipset | Example GPUs |
| -------------------------------------------- | ------------------------------------------------------------------------ |
| gfx1030 (RDNA2) | Radeon RX 6800 / 6800 XT / 6900 XT / 6950 XT, Radeon PRO W6800 |
| gfx1100, gfx1101, gfx1102 (RDNA3) | Radeon RX 7900 / 7800 / 7700 / 7600 series, Radeon PRO W7000 series |
| gfx1103 (RDNA3 integrated) | Radeon 780M / 760M in Ryzen 7040, 8040, 8000G, and 200 series processors |
| gfx1150, gfx1151, gfx1152, gfx1153 (RDNA3.5) | Integrated GPUs in Ryzen AI 300 / 400 series and Ryzen AI Max processors |
| gfx1200, gfx1201 (RDNA4) | Radeon RX 9060 / 9070 series, Radeon AI PRO R9700 |
Other RDNA2 GPUs (gfx1031 through gfx1036), such as the Radeon RX 6700 / 6600 / 6500 series and the Radeon 680M / 660M integrated GPUs, run on the gfx1030 support by [overriding the chipset](#docker-settings-for-overriding-the-gpu-chipset) to `10.3.0`.
Older GPUs, including RDNA1 (Radeon RX 5000 series) and Vega based dedicated and integrated GPUs, are not supported.
### Docker settings for GPU access
ROCm needs access to the `/dev/kfd` and `/dev/dri` devices. When docker or frigate is not run under root then also `video` (and possibly `render` and `ssl/_ssl`) groups should be added.
@@ -420,19 +436,10 @@ For reference on recommended settings see [running ROCm/pytorch in Docker](https
### Docker settings for overriding the GPU chipset
Your GPU might work just fine without any special configuration but in many cases they need manual settings. AMD/ROCm software stack comes with a limited set of GPU drivers and for newer or missing models you will have to override the chipset version to an older/generic version to get things working.
Also AMD/ROCm does not "officially" support integrated GPUs. It still does work with most of them just fine but requires special settings. One has to configure the `HSA_OVERRIDE_GFX_VERSION` environment variable. See the [ROCm bug report](https://github.com/ROCm/ROCm/issues/1743) for context and examples.
For the rocm frigate build there is some automatic detection:
- gfx1031 -> 10.3.0
- gfx1103 -> 11.0.0
If you have something else you might need to override the `HSA_OVERRIDE_GFX_VERSION` at Docker launch. Suppose the version you want is `10.0.0`, then you should configure it from command line as:
A GPU whose chipset isn't in the [supported GPUs](#supported-gpus-rocm) list can run on the support for a closely related chipset by setting the `HSA_OVERRIDE_GFX_VERSION` environment variable. RDNA2 GPUs other than gfx1030 (gfx1031 through gfx1036) need this, using `10.3.0`:
```bash
$ docker run -e HSA_OVERRIDE_GFX_VERSION=10.0.0 \
$ docker run -e HSA_OVERRIDE_GFX_VERSION=10.3.0 \
...
```
@@ -443,16 +450,17 @@ services:
frigate:
...
environment:
HSA_OVERRIDE_GFX_VERSION: "10.0.0"
HSA_OVERRIDE_GFX_VERSION: "10.3.0"
```
Figuring out what version you need can be complicated as you can't tell the chipset name and driver from the AMD brand name.
GPUs in the supported list don't need an override. An override that was set for the Radeon 780M / 760M (gfx1103) on an earlier version of Frigate is no longer needed and can be removed.
- first make sure that rocm environment is running properly by running `/opt/rocm/bin/rocminfo` in the frigate container -- it should list both the CPU and the GPU with their properties
- find the chipset version you have (gfxNNN) from the output of the `rocminfo` (see below)
- use a search engine to query what `HSA_OVERRIDE_GFX_VERSION` you need for the given gfx name ("gfxNNN ROCm HSA_OVERRIDE_GFX_VERSION")
- override the `HSA_OVERRIDE_GFX_VERSION` with relevant value
- if things are not working check the frigate docker logs
You can't tell the chipset from the AMD brand name, so check it in the container:
- first make sure that the ROCm environment is running properly by running `/opt/rocm/bin/rocminfo` in the Frigate container. It should list both the CPU and the GPU with their properties
- find the chipset version you have (gfxNNN) from the output of `rocminfo` (see below)
- if the chipset isn't in the supported list, set `HSA_OVERRIDE_GFX_VERSION` to the closest supported chipset, written as major.minor.stepping (for example gfx1030 is `10.3.0`)
- if things are not working check the Frigate docker logs
#### Figuring out if AMD/ROCm is working and found your GPU
docs/docs/frigate/hardware.md
+2 -2
View File
@@ -72,7 +72,7 @@ Frigate supports multiple different detectors that work on different types of ha
**AMD**
- [ROCm](#rocm---amd-gpu): ROCm can run on AMD Discrete GPUs to provide efficient object detection
- [ROCm](#rocm---amd-gpu): ROCm can run on AMD discrete and recent integrated GPUs to provide efficient object detection
- [Supports limited model architectures](../../configuration/object_detectors#amdrocm-gpu-detector)
- Runs best on discrete AMD GPUs
- <CommunityBadge /> [XDNA2 (Ryzen AI)](#amd-xdna2): AMD XDNA2 NPU (sub-watt power AI/ML processor separate to the GPU) inside Strix and other "AI" branded AMD platforms
@@ -240,7 +240,7 @@ Apple Silicon can not run within a container, so a ZMQ proxy is utilized to comm
### ROCm - AMD GPU
With the [ROCm](../configuration/object_detectors.md#amdrocm-gpu-detector) detector Frigate can take advantage of many discrete AMD GPUs.
With the [ROCm](../configuration/object_detectors.md#amdrocm-gpu-detector) detector Frigate can take advantage of many AMD discrete and integrated GPUs, see the [supported GPUs](../configuration/object_detectors.md#supported-gpus-rocm).
| Name | YOLOv9 Inference Time | YOLO-NAS Inference Time | RF-DETR Inference Time |
| -------------- | --------------------------- | ------------------------- | ---------------------- |