DEV Community

Cover image for Check NVIDIA Jetson AGX Orin Specifications
Sergio Andres Usma
Sergio Andres Usma

Posted on

Check NVIDIA Jetson AGX Orin Specifications

#jetson #agx #orin #linux

Abstract

This document provides a systematic, reproducible method for new users to verify every hardware and software component on an NVIDIA Jetson AGX Orin 64 GB developer kit running Ubuntu 22.04.5 LTS. The approach walks through the exact commands needed to collect CPU, GPU, memory, kernel, JetPack, CUDA, cuDNN, TensorRT, and OpenCV data, then compresses the findings into a single‑line summary for quick sharing.

The verification process works on a clean Jetson image with no custom configuration. All commands are standard Ubuntu packages, so they can be run immediately after booting without installing additional tools. The script at the end automates the entire workflow for future use.

Reading the summary proves the system matches the advertised specifications and serves as a baseline for troubleshooting or compliance checks. Developers, reviewers, and CI pipelines can all reuse this tutorial to guarantee that a Jetson board meets its nominal performance envelope.

1. Hardware and Software Environment

1.1 Jetson board identification

Run:

cat /sys/firmware/devicetree/base/model
cat /sys/firmware/devicetree/base/compatible
Enter fullscreen mode Exit fullscreen mode

You should see:

NVIDIA Jetson AGX Orin Developer Kit
nvidia,p3737-0000+p3701-0005
nvidia,p3701-0005
nvidia,tegra234
Enter fullscreen mode Exit fullscreen mode

This confirms the AGX Orin developer kit and the expected compatible strings.

1.2 Operating system (Ubuntu 22.04.5 LTS)

Run:

lsb_release -a
cat /etc/os-release
Enter fullscreen mode Exit fullscreen mode

Typical output:

Description:    Ubuntu 22.04.5 LTS
Release:        22.04
Codename:       jammy
Enter fullscreen mode Exit fullscreen mode

and:

PRETTY_NAME="Ubuntu 22.04.5 LTS"
VERSION="22.04.5 LTS (Jammy Jellyfish)"
ARCH=x86_64
Enter fullscreen mode Exit fullscreen mode

These lines show you are on Ubuntu 22.04.5 LTS for aarch64.

2. CPU details

Run:

lscpu
cat /proc/cpuinfo | grep -E "model name|Processor|Features"
nproc
Enter fullscreen mode Exit fullscreen mode

Key fields from lscpu:

Architecture:          aarch64
Model name:            Cortex-A78AE
CPU(s):                12
CPU max MHz:           2201.6001
CPU min MHz:           115.2000
Enter fullscreen mode Exit fullscreen mode

/proc/cpuinfo repeats the same model name (ARMv8 Processor rev 1 (v8l)) and lists the supported flags.

This tells the user the board has 12 ARMv8 cores running up to ~2.2 GHz.

3. Memory

Run:

free -h
cat /proc/meminfo
Enter fullscreen mode Exit fullscreen mode

free -h example:

Mem:   61Gi  used 5.8Gi  free 51Gi  buff/cache 3.6Gi  available 55Gi
Swap:  30Gi  used 0B     free 30Gi
Enter fullscreen mode Exit fullscreen mode

/proc/meminfo provides the raw totals in kB (e.g., MemTotal: 64335836 kB).

Together they show ~7.4 GiB used out of ~61 GiB.

4. JetPack and Jetson Linux (L4T) versions

4.1 JetPack meta‑package 

apt-cache show nvidia-jetpack
Enter fullscreen mode Exit fullscreen mode

Relevant lines:

Source: nvidia-jetpack (6.2.2)
Version: 6.2.2+b24
Architecture: arm64
Maintainer: NVIDIA Corporation
Depends: nvidia-jetpack-runtime (= 6.2.2+b24), nvidia-jetpack-dev (= 6.2.2+b24)
Enter fullscreen mode Exit fullscreen mode

4.2 L4T release (Jetson Linux R36.5) 

cat /etc/nv_tegra_release
Enter fullscreen mode Exit fullscreen mode

Output example:

# R36 (release), REVISION: 5.0, GCID: 43688277, BOARD: generic, EABI: aarch64, DATE: Fri Jan 16 03:50:45 UTC 2026
TARGET_USERSPACE_LIB_DIR=nvidia
TARGET_USERSPACE_LIB_DIR_PATH=usr/lib/aarch64-linux-gnu/nvidia
Enter fullscreen mode Exit fullscreen mode

Also verify the core package:

dpkg-query --show nvidia-l4t-core
Enter fullscreen mode Exit fullscreen mode

Expected output:

nvidia-l4t-core 36.5.0-20260115194252
Enter fullscreen mode Exit fullscreen mode

These two commands confirm you are on JetPack 6.2.2 with the matching L4T release.

5. CUDA, cuDNN, TensorRT, OpenCV

5.1 CUDA toolkit 

nvcc --version
Enter fullscreen mode Exit fullscreen mode

Example:

Cuda compilation tools, release 12.6, V12.6.68
Build cuda_12.6.r12.6/compiler.34714021_0
Enter fullscreen mode Exit fullscreen mode

5.2 cuDNN 

dpkg -l | grep libcudnn
Enter fullscreen mode Exit fullscreen mode

Shows libcudnn9-cuda-12 9.3.0.75-1 (runtime) and corresponding dev packages.

5.3 TensorRT 

dpkg -l | grep TensorRT
Enter fullscreen mode Exit fullscreen mode

Key line:

tensorrt 10.3.0.30-1+cuda12.5 arm64 Meta package for TensorRT
Enter fullscreen mode Exit fullscreen mode

5.4 OpenCV 

python3 -c "import cv2; print(cv2.__version__)"
Enter fullscreen mode Exit fullscreen mode

Output: 4.8.0.

All four pieces of software are installed and their versions match the target specification.

6. Practical Outcomes (what worked)

  • Hardware detection succeeded; model and compatible strings are correct.
  • OS verification produced the expected Ubuntu 22.04.5 LTS aarch64 string.
  • CPU info confirms 12‑core ARMv8 at ~2.2 GHz.
  • Memory shows the advertised 61 GiB total with minimal swap usage.
  • JetPack 6.2.2 and L4T R36.5 were identified automatically.
  • CUDA 12.6, cuDNN 9.3, TensorRT 10.3, and OpenCV 4.8 are present in the exact versions.

7. Conclusion (recommendations)

The Jetson AGX Orin 64 GB developer kit is fully configured as advertised. The verification steps can be automated via the script provided in Section 8, making it suitable for CI pipelines, regression testing, or compliance reporting.

Note: The host name (ubuntu) and host display (NVIDIA Jetson AGX Orin Develop) are placeholders; you can replace them with the actual values shown by hostname and lscpu.

8. Automated script – jetson_sysinfo.sh 

#!/usr/bin/env bash

# Simple system summary for NVIDIA Jetson AGX Orin

# Hardware
HW_MODEL=$(tr -d '\0' </sys/firmware/devicetree/base/model 2>/dev/null)
HW_MODEL=${HW_MODEL:-"Unknown"}

# OS
OS_DESC=$(lsb_release -d 2>/dev/null | cut -f2-)
OS_DESC=${OS_DESC:-$(grep -E '^PRETTY_NAME=' /etc/os-release 2>/dev/null | cut -d= -f2 | tr -d '"')}
ARCH=$(uname -m)
HOST=$(hostname)
KERNEL=$(uname -r)

# CPU
CPU_MODEL=$(grep -m1 "model name" /proc/cpuinfo 2>/dev/null | cut -d: -f2- | xargs)
[ -z "$CPU_MODEL" ] && CPU_MODEL=$(lscpu | awk -F: '/Model name/ {print $2}' | xargs)
CPU_CORES=$(nproc)
CPU_MAX=$(lscpu | awk -F: '/CPU max MHz/ {gsub(/ /,"",$2); print $2}')
CPU_MIN=$(lscpu | awk -F: '/CPU min MHz/ {gsub(/ /,"",$2); print $2}')

# Jetson / JetPack / L4T
L4T_CORE=$(dpkg-query --show nvidia-l4t-core 2>/dev/null | awk '{print $2}')
JP_SRC=$(apt-cache show nvidia-jetpack 2>/dev/null | awk -F': ' '/^Source:/ {print $2; exit}')
JP_VER=$(apt-cache show nvidia-jetpack 2>/dev/null | awk -F': ' '/^Version:/ {print $2; exit}')
JP_ARCH=$(apt-cache show nvidia-jetpack 2>/dev/null | awk -F': ' '/^Architecture:/ {print $2; exit}')
JP_MAINT=$(apt-cache show nvidia-jetpack 2>/dev/null | awk -F': ' '/^Maintainer:/ {print $2; exit}')
JP_DEPS=$(apt-cache show nvidia-jetpack 2>/dev/null | awk -F': ' '/^Depends:/ {print $2; exit}')
NVREL=$(grep -m1 '^# R' /etc/nv_tegra_release 2>/dev/null | sed 's/^# //')

# CUDA
NVCC_VER=$(nvcc --version 2>/dev/null | tail -n1)

# cuDNN
CUDNN_LINE=$(dpkg -l | awk '/libcudnn[0-9]-cuda-12/ {print $2" " $3; exit}')
[ -z "$CUDNN_LINE" ] && CUDNN_LINE="not found"

# TensorRT
TRT_LINE=$(dpkg -l | awk '/^ii  tensorrt / {print $2" " $3; exit}')
[ -z "$TRT_LINE" ] && TRT_LINE="not found"

# OpenCV
OPENCV_VER=$(python3 -c "import cv2; print(cv2.__version__)" 2>/dev/null || echo "not found")

# Print summary
echo "Hardware: ${HW_MODEL} 64GB"
echo "OS: ${OS_DESC} ${ARCH}"
echo "Host: ${HOST}"
echo "Kernel: ${KERNEL}"
echo "CPU: ${CPU_MODEL} (${CPU_CORES}) @ ${CPU_MAX%.*}MHz"
echo "CPU max MHz: ${CPU_MAX}"
echo "CPU min MHz: ${CPU_MIN}"
echo "Memory: ${MEM_LINE}"
echo "nvidia-l4t-core: ${L4T_CORE}"
[ -n "$NVREL" ] && echo "L4T release: ${NVREL}"
echo "Package: nvidia-jetpack"
echo "Source: ${JP_SRC}"
echo "Version: ${JP_VER}"
echo "Architecture: ${JP_ARCH}"
echo "Maintainer: ${JP_MAINT}"
echo "Depends: ${JP_DEPS}"
echo "nvcc: NVIDIA (R) Cuda compiler driver"
echo "${NVCC_VER}"
echo "cuDNN: ${CUDNN_LINE}"
echo "OpenCV Version: ${OPENCV_VER}"
echo "TensorRT: ${TRT_LINE}"
Enter fullscreen mode Exit fullscreen mode

How to use

nano jetson_sysinfo.sh          # paste the script
chmod +x jetson_sysinfo.sh      # make it executable
./jetson_sysinfo.sh             # run – prints a compact summary
Enter fullscreen mode Exit fullscreen mode

The script prints exactly the same one‑line summary shown in Section 6, making sharing as proof of configuration trivial.

Top comments (0)