Recently, I upgraded my Jetson Orin Nano from JetPack 5.X to the latest JetPack 6.2. This represents a significant update, moving from Ubuntu 20.04 to Ubuntu 22.04 as the base OS and bringing numerous performance improvements. I’ve documented the entire process to help others make this transition smoothly.
Why Upgrade?
JetPack 6.1 offers several compelling benefits:
- Improved AI framework support
- Better performance for inference workloads
- Enhanced CUDA optimization
- Latest security updates
- Support for newer libraries and tools
What You’ll Need
- NVIDIA SDK Manager (on a host computer)
- NVMe SSD
- USB Type-C cable to connect to your host computer
- Reliable power supply for your Jetson Orin Nano
Step 1: Download NVIDIA SDK Manager
- Visit the NVIDIA Developer website
- Download the latest version compatible with your host computer
- Install and launch the SDK Manager
I recommend installing Ubuntu 20.04 for the smooth upgrade.
How to connect Jetson Orin to Ubuntu Host
Use a cable that connects Type C USB port of the Jetson Orin to the Type A USB of the host system.
Step 2: Prepare for the Upgrade
Before proceeding, it’s crucial to back up all your important data:
Step 1. Back up your data
Save any important projects, configurations, or user data before you perform the upgrade. Export any Docker containers or environments you want to preserve.
Step 2. Ensure that NVMe gets detected
For AI workload, I always recommend using NVMe SSD. Once you install SDK Manager on the Ubuntu system, you should be able to detect NVMe SSD card.
Step 3: Flash JetPack 6.X Using SDK Manager
- In SDK Manager, select “Jetson” under “Target Platform”
- Select your Jetson Orin Nano model
- Choose “JetPack 6.X” from the available versions
- Select the components you wish to install (CUDA, TensorRT, etc.)
- Click “Continue” to proceed to the installation step
Step 5: Complete the Flashing Process
- In SDK Manager, follow the prompts to install JetPack 6.X
- Choose “Automatic Setup” for the simplest experience
- SDK Manager will download components and flash your device
- This process may take 30-45 minutes depending on your internet speed
Step 6: Initial Setup and Verification
- After flashing completes, your Jetson will reboot
- Complete the first-boot setup (user account, network, etc.)
- Verify the JetPack version:
cat /etc/nv_tegra_release
Look for R36, which corresponds to JetPack 6.X.
cat /etc/nv_tegra_release
# R36 (release), REVISION: 4.3, GCID: 38968081, BOARD: generic, EABI: aarch64, DATE: Wed Jan 8 01:49:37 UTC 2025
# KERNEL_VARIANT: oot
TARGET_USERSPACE_LIB_DIR=nvidia
TARGET_USERSPACE_LIB_DIR_PATH=usr/lib/aarch64-linux-gnu/nvidiaStep 7: Optimize Your System
Get the most out of your newly upgraded system:
Set the desired performance mode:
sudo nvpmodel -m 0 # MAXN mode for maximum performance
sudo jetson_clocks # Set maximum clock speeds
Install any additional software you need:
sudo apt update
sudo apt install nvidia-jetpack
Performance Improvements
After upgrading to JetPack 6.X, I noticed significant performance improvements in my AI workloads. Models that previously ran at 5-7 FPS now achieve 8-10 FPS, representing a 40-60% improvement in some cases.
Troubleshooting Tips
If you encounter issues with WiFi or Bluetooth after the upgrade, try:
sudo apt install --reinstall nvidia-l4t-wifibt
For graphics issues:
sudo apt install --reinstall nvidia-l4t-core nvidia-l4t-graphics-drivers
If certain applications stop working, check their compatibility with Ubuntu 22.04
Important Considerations
- JetPack 6.X uses Ubuntu 22.04 as the base OS (JetPack 5.x used Ubuntu 20.04)
- Some custom applications may need recompiling for Ubuntu 22.04
- Hardware peripherals/drivers may require updates for full compatibility
Conclusion
Upgrading to JetPack 6.X breathes new life into the Jetson Orin Nano, especially for AI and computer vision applications. The process requires a complete system reinstallation, but the performance benefits make it well worth the effort.
Have you upgraded your Jetson device? Share your experience in the comments below!
