how can you tell if a gpu would work with frigate? im not quite sure the requirements for facial recognition and other things.

[JJandPenguin]

how can you tell if a gpu would work with frigate? im not quite sure the requirements for facial recognition and other things.
i have a visiontek 1gb radeon hd7750 sff gddr3 for free that my friends going to give me and im curios if it will work.

[inkeep-ai-assistant[bot]]

---

For GPU compatibility with Frigate, you need to check what hardware is officially supported for different tasks.

Object Detection

For object detection, Frigate supports specific GPU types(1):

• AMD GPUs: ROCm can run on AMD Discrete GPUs to provide efficient object detection(1)
• Intel: OpenVino can run on Intel Arc GPUs, Intel integrated GPUs, and Intel CPUs(1)
• Nvidia GPU: Nvidia GPUs will automatically be detected and used as a detector in the -tensorrt Frigate image(1)

Your AMD Radeon HD 7750 is quite old (released in 2012). ROCm support for AMD GPUs is limited to more recent architectures. Based on the documentation, ROCm requires specific GPU chipset configurations(1), and older cards like the HD 7750 are unlikely to be supported.

Face Recognition

For face recognition specifically, Frigate offers two model types(2):

• small: Runs locally on the CPU using a FaceNet embedding model, optimized for efficiency(2)
• large: Runs an ArcFace embedding model optimized for accuracy, and it is only recommended when an integrated or dedicated GPU/NPU is available(2)

The minimum system requirements state that the large model "is optimized for accuracy, an integrated or discrete GPU / NPU is required"(2).

Hardware Acceleration for Video Decoding

Hardware acceleration is primarily used for video decoding (not object detection when using dedicated AI accelerators like Coral)(3). The supported hwaccel presets would need to match your GPU capabilities(3).

Your Specific GPU

Unfortunately, the Radeon HD 7750 is from 2012 and uses the GCN 1.0 architecture. ROCm typically supports GCN 4.0 and newer (Polaris/Vega/RDNA architectures). Your card is unlikely to work with Frigate's ROCm-based object detection or the large face recognition model.

You could potentially use it for basic video decode acceleration if the drivers support VA-API, but you wouldn't get GPU-accelerated AI inference for object detection or face recognition.

📚 Sources:

1. Supported Hardware
2. Face Recognition
3. High CPU Usage

---

Have another question? Just tag @inkeep.

[inkeep-ai-assistant[bot]]

---

For Nvidia GPUs with Frigate, the documentation states that "Nvidia GPUs will automatically be detected and used as a detector in the -tensorrt Frigate image when a supported ONNX model is configured"(1).

Your Nvidia Tesla K80

The Tesla K80 is based on the Kepler architecture (GK210 GPU, released in 2014). While it's technically an Nvidia GPU, there are important considerations:

TensorRT Support: The K80 uses CUDA compute capability 3.7. Modern TensorRT versions typically require compute capability 5.0 or higher (Maxwell architecture and newer). This means the K80 may not be compatible with current TensorRT implementations in Frigate.

ONNX Detector Configuration: According to the documentation, when using the -tensorrt image, you can configure multiple detectors if GPU resources are available(1):

detectors:
  onnx_0:
    type: onnx
  onnx_1:
    type: onnx

(1)

Hardware Acceleration for Video Decoding: The K80 could potentially be used for hardware-accelerated video decoding, which "significantly reduce CPU usage per camera stream" and allows you to "support 2-3x more cameras on the same hardware"(2). However, you would need to configure hwaccel_args based on your hardware(2).

Bottom Line

The Tesla K80 is likely too old for Frigate's TensorRT-based object detection due to its older CUDA compute capability. You might get some benefit for video decode acceleration, but it won't provide GPU-accelerated AI inference for object detection or face recognition tasks. A more modern GPU (GTX 1050 or newer) would be recommended for full Frigate GPU acceleration.

📚 Sources:

1. Supported Hardware
2. High CPU Usage

---

Have another question? Just tag @inkeep.

[JJandPenguin]

@inkeep what would be a good older / cheaper gpu for frigate, i only need 2-4 cameras

[inkeep-ai-assistant[bot]]

Based on the available documentation, I don't have specific GPU model recommendations for budget hardware. However, I can share what hardware Frigate officially supports for object detection(1):

Supported GPU Options

Intel GPUs - OpenVINO can run on Intel Arc GPUs, Intel integrated GPUs, and Intel CPUs(1). OpenVINO is supported on 6th Gen Intel platforms (Skylake) and newer(1).

AMD GPUs - ROCm can run on AMD Discrete GPUs to provide efficient object detection(1). However, ROCm typically requires newer AMD architectures and your GPU chipset may need manual configuration(1).

Nvidia GPUs - Nvidia GPUs will automatically be detected and used as a detector in the -tensorrt Frigate image when a supported ONNX model is configured(1).

Alternative: Google Coral Edge TPU

For 2-4 cameras, you might want to consider the Google Coral Edge TPU instead of a GPU. The Coral EdgeTPU is available in USB, Mini PCIe, and m.2 formats allowing for a wide range of compatibility with devices(1). These are specifically designed for efficient object detection and are often more cost-effective than GPUs for this purpose.

Configuration Note

If you have both NPU and GPU available (Intel Core Ultra processors), the documentation recommends using NPU for object detection and GPU for enrichments (semantic search, face recognition, etc.) for best performance and compatibility(1).

For specific older GPU model recommendations within your budget, I'd suggest checking the Frigate community forums or GitHub discussions where users share their hardware experiences.

📚 Sources:

1. Supported Hardware

---

Have another question? Just tag @inkeep.