Skip to main content

NVIDIA Jetson Orin Nano Developer Kit

Current Status
SupportedSDKProvisioning
🟢 Target actively maintained🟢 x86-64 and aarch64 container images🟡 NVMe storage

Jetson Orin Nano Developer Kit

The Jetson Orin Nano Developer Kit delivers exceptional performance for real-time ML at the edge—up to 67 TOPS of AI compute. Paired with Avocado OS, you can deploy full inference pipelines in minutes, without the typical headaches of cross-compiling or system reboots.

Whether you're building computer vision, robotics, or edge AI applications, this target gets you production-ready fast.

🛠 Technical Specifications​

ComponentDetails
CPU6-core Arm Cortex-A78AE v8.2 (1.7 GHz)
GPUNVIDIA Ampere GPU with 1024 CUDA cores and 32 tensor cores
AI PerformanceUp to 67 TOPS (INT8)
Memory8GB 128-bit LPDDR5
Memory Bandwidth102 GB/s
Storage2 x M.2 Key M slots for PCIe NVMe SSDs
ConnectivitySingle M.2 Key E wireless module with Wi-Fi and Bluetooth
Power Modes7W / 15W / 25W

🚀 Getting Started​

Get up and running with the Avocado Linux SDK in minutes.

Prerequisites​

  • A Mac (macOS 10.12+) or Linux (Ubuntu 22.04+, Fedora 39+) development machine
  • Docker installed
  • 10GB+ available disk space

Installing and running the SDK​

  1. Install the Avocado CLI and append the location of the avocado executable to your PATH environmant variable.

  2. Create your project workspace:

mkdir avocado-jetson-orin-nano
  1. Initialize a new project:
cd avocado-jetson-orin-nano
avocado init
  1. Replace qemux86-64 with jetson-orin-nano-devkit-nvme as target in avocado.toml:
[runtime.dev]
target = "jetson-orin-nano-devkit-nvme"
  1. Install all components (SDK, extensions, and runtime dependencies):
avocado install -f

⚙️ Provisioning​

We are actively working on a provisioning guide for the Jetson Orin Nano Developer Kit.

🧰 Hardware-in-the-Loop (HIL)​

We are actively working on Hardware-in-the-Loop (HIL) development for the Jetson Orin Nano Developer Kit.

🤖 Deploying ML Inference with Triton​

With Avocado OS, you can deploy NVIDIA’s Triton Inference Server in just six commands—no cross-compiling or reflashing required.

Why it matters:

  • Model updates apply live, without device reboots or service restarts
  • Works seamlessly with Avocado’s OTA update infrastructure
  • Supports Hardware-in-the-Loop (HIL) testing workflows

See how we built this at Open Source Summit →

🧭 Target Roadmap/Known Limitations​

🟢 GPU-accelerated ML inference is fully supported
🟡 GPU-accelerated video pipelines are under development
🟡 NVMe provisioning is under development
🟡 Hardware-in-the-Loop (HIL) debugging is under development
🟠 Secure boot is not yet supported
🟠 Full disk encryption is not yet supported