How to Add More Storage to NVIDIA Jetson Projects

The NVIDIA Jetson platform continues to push the boundaries of Edge AI, providing solutions with accelerated capabilities in very small form factors. If your Jetson project could benefit from additional storage, here are a few options to choose from.

Developer Kits vs. Production Modules

The NVIDIA Jetson Developer Kit is a great way to begin your development. When working with the Jetson Nano or Xavier NX Developer Kit, the option to boot from expandable SD storage is built into the Kit. Both the Nano and Xavier NX production modules come equipped with 16GB of on-board eMMC storage, but does not include the integrated SD card slot. If your project is data extensive, you may require a carrier board that provides additional storage options. Considering when the latest JetPack and SDK application components are fully loaded, they can use up to 14GB of storage – additional storage flexibility could be invaluable to your project. Here is a comparable list of the Jetson storage capabilities of the developer kit and production modules:

	Nano	TX2	Xavier NX	AGX Xavier
Developer Kit	SD Card NVMe	32GB eMMC SD Card NVMe	microSD NVMe	32GB eMMC NVMe SD Card
Production Module	16GB eMMC	32GB eMMC	16GB eMMC	32GB eMMC

Due to the data payload of high-end sensors, many applications will need a solution to store application, video, and sensor data when progressing to production.

Storage Decisions

Before selecting hardware for production, you must first decide what storage solution is right for you. If the on-board eMMC storage within the module itself is sufficient for your project, this process is easy. Projects requiring integration of multiple sensors or high data payloads will benefit from having storage flexibility beyond the provided eMMC storage. In most cases, you’ll need to decide if you want to store applications on an SD card or through an NVMe SSD. What’s right for your project?

SD Card	NVMe
PROs • Readily-available and inexpensive • Up to 2TB of storage	PROs • Fastest possible load time • Increased reliability • Up to 2TB of storage
CONs • Possible loading delays • Can be unreliable	CONs • More expensive compared to SD cards

Choosing a Carrier Board with SD or NVMe Pre-integrated

NVIDIA production modules ship without the Dev Kit’s reference carrier board, making it necessary to find a carrier board that fits your project’s unique requirements. Additional storage requirements means finding a carrier board that has an NVMe or SD slot available. If the selected board has your required I/O connections and matches size requirements, moving your application from the NVIDIA Dev Kit to the chosen carrier board for deployment is a straight forward process. If you can’t find something out-of-the-box, you may need to venture down the path of modifying an existing board or completely customizing a carrier board for your project.

Booting Applications off NVMe or SD Cards

Ideally, the application and sensor data will all be housed within the same storage device. Storing application data on a different drive than sensor data will cause unnecessary lag – which is unacceptable in most modern AI applications.

There is a simple solution that can be configured when flashing the Jetson module that will allow you to boot programs off an SD card or NVMe. For Nano and Xavier NX users, follow these simple instructions.

If you’re looking for information on how to boot off NVMe using an AGX Xavier module, these Jetson Developer Forum users have outlined the steps – view them here.