For a first Jetson project, the Orin Nano Super Developer Kit gives you an 8GB development computer with a reference carrier board. Consider an Orin NX module when your measured memory, accelerator or processing requirements exceed that setup, and you can provide a compatible carrier, storage, cooling and power supply. Choose the exact product and software configuration before comparing prices or AI performance numbers.
This guide compares the existing Orin Nano and Orin NX families. NVIDIA’s August 25, 2026 Orin Nano 2 announcement describes a separate module and developer kit expected in the first half of 2027. That is a company availability plan, not a product used or purchased for this guide. Sources below were checked October 14, 2026; no Jetson, camera or robot benchmark was run here.
A module is one part of a working computer
A production module contains the processor and memory. A developer kit combines a development module with a carrier board and other development hardware. A third-party finished system can add a different carrier, enclosure, storage and interfaces. Three listings bearing “Orin” can therefore include different parts.
The NVIDIA Orin comparison lists the Nano Super kit separately from production modules. Its reference carrier supports Orin Nano and Orin NX modules, but this does not make every third-party carrier interchangeable. Match the exact module, carrier revision and board-support package.
For a supplier quote, write down the module model and memory, carrier, storage, thermal solution, power supply and included software. Ask which components are absent. A module price is not the price of a complete robot computer.
Compare memory and supported modes before TOPS
NVIDIA’s current specifications give the following Super configuration values. These are manufacturer maximum sparse INT8 AI-throughput figures and selectable module power ranges, not our application measurements.
| Product | Memory | Up to sparse INT8 TOPS | Module power modes |
|---|---|---|---|
| Orin Nano Super Developer Kit | 8GB | 67 | 7–25W |
| Orin NX 8GB module | 8GB | 117 | 10–40W |
| Orin NX 16GB module | 16GB | 157 | 10–40W |
The specification and compute tables identify the calculation type. A TOPS ratio is not a frame-rate, language-model latency or robot success ratio. Power mode, software, precision and the actual operations supported by an accelerator affect the result. The older figures in the same page’s family descriptions should not be mixed with these Super configuration values.
The distinction between the two NX models matters. Moving from Nano 8GB to NX 8GB adds compute features, but not a larger advertised memory capacity. If profiling your intended process set shows that its resident allocation exceeds 8GB, a faster 8GB module does not resolve that capacity requirement. NX 16GB is a candidate to evaluate; it still needs a measurement with your full workload. CPU processes and GPU work share the module’s memory budget.
Make the camera and connector plan concrete
The Nano developer-kit hardware layout shows two 22-pin CSI connectors, four USB Type-A ports, Gigabit Ethernet and M.2 storage slots. Its display output is DisplayPort; the USB-C connector does not provide a display signal. A monitor with only HDMI needs the appropriate DisplayPort-to-HDMI connection. A 15-pin CSI camera needs the correct 15-to-22-pin cable as well as software support.
For each proposed camera, identify its interface, connector, sensor or device model, driver and supported Jetson Linux release. Also check how its image format, resolution and timestamps reach your application. A matching plug does not establish that the driver, simultaneous streams or synchronization will work.
For example, a project using two USB cameras and an NVMe SSD should confirm the carrier’s available ports, USB topology and storage configuration before comparing GPUs. A design using serialized CSI cameras needs a different carrier and driver investigation. The module’s theoretical camera lanes are not a count of ready-to-use connectors on every kit.
Choose software and cooling as part of the purchase
The current JetPack download notes list JetPack 7.2.1, Jetson Linux 39.2.1 and Orin-family support. They describe a unified ISO flashing path and explicitly state that there is no longer a Nano developer-kit SD-card image for that release. Earlier JetPack 6 tutorials can therefore describe a different installation path. The same current notes label Isaac ROS support “Coming soon”; do not assume that choosing the newest JetPack also supplies the Isaac ROS stack your project requires.
Confirm a software combination supported by your carrier, camera drivers and robot packages. Record the JetPack, Jetson Linux and inference-library versions together. Our TensorRT engine-reuse guide explains why moving a saved engine between boards is a separate compatibility question.
The module’s listed power range is not your whole system’s battery or power-supply budget. Cameras, carrier, storage and cooling add loads. Test your chosen mode under the intended sustained process set and ambient conditions. The sustained edge-compute guide explains how to investigate throttling and application latency rather than relying on a short peak result.
Orin Nano Super is the original kit with a faster power mode
If you searched for “Jetson Orin Nano Super”, you are looking at the same developer kit hardware with a new top power mode. NVIDIA’s December 17, 2024 launch post renamed the kit after a software update raised its clocks. In that mode NVIDIA lists:
- 67 sparse INT8 TOPS, up from 40
- 1,020 MHz GPU clock, up from 635 MHz
- 1.7 GHz CPU clock, up from 1.5 GHz
- 102 GB/s memory bandwidth on the same 8GB LPDDR5
- a 25W module power mode added to the existing 7W and 15W modes
NVIDIA measured 1.28x to 1.63x more tokens per second on the small language models it tested, such as Llama 3.2 3B, in that post. Those are NVIDIA’s own figures with specific quantization settings, not a result for your robot pipeline.
If you already own the original Orin Nano Developer Kit, you do not need new hardware. NVIDIA says every earlier kit can use the faster mode after moving to JetPack 6.1 or later:
- Flash the current SD card image from the JetPack downloads page, or install JetPack with an updated SDK Manager.
- Boot the kit and run
sudo nvpmodel -m 2, or pick the MAXN mode in the power selector on the Ubuntu top bar. - Plan for the higher power draw: use the supplied power adapter and keep the fan working, because sustained 25W loads heat the module more than the 15W mode. Our sustained thermal test guide shows how to check for throttling.
NVIDIA extended the same “Super” clocks to Orin Nano and Orin NX production modules through later JetPack releases, so a module listed as “Super” in a third-party system is usually the same silicon running a newer software configuration. Confirm the power mode and JetPack version in the seller’s specification before comparing numbers.
Use a price reference without treating it as a current quote
NVIDIA’s December 17, 2024 Super launch article announced a $249 developer-kit price. After NVIDIA’s July 2026 price increase, the Jetson FAQ lists the Orin Nano Super Developer Kit at $399 (checked October 20, 2026). The $249 figure is a dated reference for the Nano Super kit, not an October 2026 retailer quote, an NX module price or a delivered system cost. Current regional stock, tax, shipping and seller bundles must be checked at purchase. We did not obtain an NX-system supplier quote for this article.
Start with Nano Super for an affordable development setup when its memory and interfaces suit the project. Move to an NX candidate for a defined compute or memory need, then compare complete-system quotes and test the resulting configuration. If the decision instead involves moving to the larger Thor platform, use the separate AGX Thor versus Orin guide. Keep that platform choice separate from selecting an Orin Nano or NX module.