An NVIDIA Orin
|
The 69.6x45 Jetson Orin module, shown on the right, is a production-ready System on Module (SOM) that is targeted at embedded industrial AI applications such as sophisticated robotics or surveillance, and would typically be part of a custom developed overall system.
This is an extremely powerful platform which has recently (Dec'24) been given a major software upgrade to support emerging applications for generative AI workloads in robotics and multimodal agents and can deliver up to 67 TOPS of AI performance.
The Orin Development Kit is a very complete set of hardware provided by NVIDIA that has a custom carrier board, a heatsink with an attached cooling fan, and provides convenient exposure of most of the system interfaces including a pair of vacant built-in NVME connectors on the underside of the board, and where the 2x20 GPIO expansion header also mimics the standard Raspberry Pi layout.
The build using this dev kit has used a custom 3D printed case (print files downloadable from here) designed using FeeCAD, with the following set of images showing various components and different stages of the assembly.
| FreeCAD image of the case base | FreeCAD image of the underside of the case lid |
FreeCAD image of the dev kit placed in the case base |
FreeCAD image of the fully assembled dev kit in its case |
| Connector-side view of the dev kit | SD-side view of the dev kit | underside of dev kit with NVME SSD installed |
dev kit placed in the case base |
| SD, button and CAN Bus headers opening on side of dev kit lid |
case lid side opening for the two MIPI CSI Camera connectors |
cable connecting through to the 2x20 pin GPIO header |
vented underside of the case base with self-adhesive feet added to improve air flow |
Some noteworthy specific features of the Orin dev kit, and some aspects of this build include:
- The dev kit carrier board only has a displayport connector for a screen connection, so for most screens a displayport-to-hdmi cable is needed.
- Running the OS from a SD card is the main method in all the NVIDIA documentation, and this build has used the recommended 128GB micro SD card.
- A 1TB NVME SSD has been installed in one of the NVME connectors on the underside of the carrier board - and this is another recommendation for using the platform for a range of generative AI uses.
The lid of the custom 3D printed case has a 'breakout' opening that allows a 2x20 cable with a reasonably wide connector to pass through the case to the GPIO header on the dev kit's carrier board. The cable can obviously be whatever length is appropriate, with an image above showing a reasonably long cable, and the image on the right showing how the RPi Maker PCB can be connected with a short cable to enable a wide range of sensors and other devices to be easily connected and tested with the Orin.
- The SD-side of the case lid has a wide opening that not only provides easy access to the SD card but also provides access to the 12-pin button header and the optional 4-pin CAN Bus header.
- The non-GPIO header side of the case lid also has an opening for two camera cables to pass through when connected to the MIPI CSI Camera connectors.
IoT hardware platform build details:
More about IoT:
All the currently available maker project information: