-
Notifications
You must be signed in to change notification settings - Fork 3
Compatible Computer Systems
Nvidia has developed a range of embedded systems specifically designed for robotics and AI applications, known as the Jetson series. These devices vary in performance, power consumption and feature support. You can find an overview of the Jetson family here.
To determine which Jetson module is best suited for use with the ZED2i camera, I evaluated each system based on the following key metrics:
- GPU: The number of CUDA cores is directly related to the system’s ability to handle real-time depth mapping, image processing, and object detection. Generally, more CUDA cores translate to better parallel processing capabilities.
- CUDA Support: Since the ZED SDK depends on Nvidia CUDA, only systems with the required CUDA compute capability can run the SDK reliably. Compatibility with newer versions of CUDA ensures better support for recent features and performance improvements.
- Power Range: Power consumption is a critical factor when integrating a Jetson system into a robot. Understanding the power requirements allows for proper hardware design.
- Performance Rating: This metric in the table is a qualitative assessment based on the GPU specs and CUDA support. This is subjective, however it reflects general trends in how well each system handles ZED2i features such as depth sensing and object detection.
| System | GPU | CUDA Support | Power Range | Performance Rating | Notes |
|---|---|---|---|---|---|
| Jetson Nano | Maxwell (128 Core) | CUDA 10.2 | 5W-10W | Low | Not Recommended for Full SDK |
| Jetson Xavier NX | Volta (384 Core + 48 Tensor) | CUDA 11.x | 10W-20W | Good | Recommended for lightweight tasks |
| Jetson AGX Xavier | Volta (512 Core + 64 Tensor) | CUDA 11.x | 10W-30W | Very Good | Recommended |
| Jetson Orin Nano | Ampere (1024 Core + 32 Tensor) | CUDA 11.x/12.x | 7W-25W | Very Good | Recommended |
| Jetson AGX Orin | Ampere (2048 Core + 64 Tensor) | CUDA 12.x | 15W-60W | Excellent | Best for high end applications |
Based on the comparison table above, the most suitable systems for use with the ZED2i camera are the Jetson AGX Xavier and the Jetson Orin Nano. Both offer strong performance, adequate CUDA support, and compatibility with the ZED SDK.
However, it’s important to note that the Jetson AGX Xavier was released in 2018 and while still powerful, it is based on an older architecture. If both options are available at similar price points, the Jetson Orin Nano is the more modern choice, offering newer architecture and longer-term software support.