1: Hardware preparation

1. In this article, we will use the SDK Manager for setup.

Connect the Jetson Orin Nano (Super) developer kit to the host PC (Linux: Ubuntu) using a USB cable and Type-C USB​ ​port with the following configuration. 

The Jetson Orin Nano (Super) Developer Kit requires a separate microSD card (64GB or more recommended) or NVMe SSD (256GB or more recommended).

*If you are running the generative AI demo, we recommend that you use an NVMe SSD (256GB or more).

2. Follow the instructions on the manufacturer's website to boot into Force Recovery Mode.

3. Connect the AC adapter included with the Jetson Orin Nano (Super) Developer Kit.

4. Enter the following command on the host PC to confirm that the USB interface of the Jetson Orin Nano(Super) Developer Kit is recognized.

host$ lsusb

Check to see if “NVIDIA Corp.” is displayed as shown below.

2: Software preparation

1. Manufacturer site Refer to the guide to host PC fart SDK Manager Install.

*Supported host PC OS requirements are Ubuntu 20.04, Ubuntu 22.04, and Ubuntu 24.04.

2. On the host PC, enter the following command to launch the SDK Manager.

*To run the SDK Manager, you must be a registered NVIDIA user.

 host$ sdkmanager

3. As shown in the image below, make sure that STEP 1 shows JetPack 6.2, then click "CONTINUE."

4. Next, in STEP 2, check Jetson SDK Components and click "CONTINUE".

5. The window below will appear during the process, so set your "Username" and "Password".

For “Storage Device”, select the SD card or NVMe SSD that you connected separately to the Jetson Orin Nano (Super) development kit.

6. When writing is complete, "FINISH" will be displayed as shown below.

In addition to the method of writing firmware using SDKManager mentioned above, there is also a writing method that does not require a host PC and uses only an SD card image, so please refer to this manufacturer's website.

3: Setting Super mode

Configure Super mode on the Jetson Orin Nano (Super) developer kit.

1. Left-click on the "Power display"in the upper right corner, select "Power mode" and select "MAXN SUPER" to switch to Super mode.

2. In the same way, select "Run Jetson Power GUI" to check the configuration of the Jetson Orin Nano (Super) Developer Kit.

　　The CPU and GPU Freq (MHz) will vary depending on the operating conditions of the Jetson, but you should be able to confirm that the upper frequency limit for Super mode is approximately 1.7GHz.

1: Jetson Orin Nano Super Developer Kit Initial Setup

Various generative AI demos for Jetson are provided in the Docker container image for Jetson​ ​(jetson-containers) provided by NVIDIA. We will configure these to use them.

The Ubuntu user is not included in the group by default, so add the user to the Docker group with the command below.

sudo apt get update
sudo usermod -aG docker $USER
newgrp docker

When building a Docker container, modify the configuration file (daemon.json) as follows to enable the use of GPUs.

sudo vi /etc/docker/daemon.json

{

    "runtimes": {

        "nvidia": {

            "path": "nvidia-container-runtime",

            "runtimeArgs": []

        }

    },

    "default-runtime": "nvidia"

}

Next, clone and set up the Docker container with the following command.

sudo systemctl daemon-reload && sudo systemctl restart docker 
git clone https://github.com/dusty-nv/jetson-containers
bash jetson-containers/install.sh

When running large models such as generative AI, the RAM capacity of the Jetson Orin Nano module is insufficient, so disable ZRAM with the following command and create a swap file. A swap file can use part of the storage space of an SSD or other device as if it were RAM when the RAM capacity is insufficient.

sudo systemctl disable nvzramconfig
sudo fallocate -l 16G /mnt/16GB.swap
sudo mkswap /mnt/16GB.swap
sudo swapon /mnt/16GB.swap

2: Running the Small Language Model (SLM) demo app

Next, run the chat app using the Small Language Model (SLM) with the following command.

jetson-containers run $(autotag nano_llm) \ 
  python3 -m nano_llm.chat --api=mlc \ 
    --model princeton-nlp/Sheared-LLaMA-2.7B-ShareGPT 
    --max-context-len=512 

When the app launches, it automatically enters interactive mode, allowing you to ask questions at a command prompt.

This time, let's try asking about travel destinations in Japan.

Also, if you check the Power GUI at this time, you can see that the GPU operating frequency is increasing to perform AI processing.

The maximum GPU operating frequency in Super mode is 1020MHz.

If you are considering introducing AI, please contact us.

For the introduction of AI, we offer a wide range of services, including the selection and support of NVIDIA GPU cards and GPU workstations, as well as algorithms for face recognition, trajectory analysis, and skeletal detection, and learning environment construction services. If you need help, please feel free to contact us.