Jetson AGX Orin 32GB on JetPack 6.2.3: NEEDS VALIDATION
Method v1.0 · dataset 2026-09-07 · verified 2026-09-07 · last updated September 2026
Jetson AGX Orin 32GB on JetPack 6.2.3 (Jetson Linux 36.5.2), SDK Manager (host flash), TensorRT: NEEDS VALIDATION. jetpack release NEEDS VALIDATION; runtime NEEDS VALIDATION. 7 verify commands settle it on the board.
Jetson AGX Orin 32GB · JetPack 6.2.3 (L4T 36.5.2) · CUDA — · TensorRT — · cuDNN — · Ubuntu 22.04
SDK Manager flash · TensorRT runtime. Change the install method, runtime or power mode on the live engine.
Checks
| Check | Status | Assumption | Evidence |
|---|---|---|---|
| jetpack release | REQUIRES_VALIDATION | JetPack 6.2.3 = Jetson Linux 36.5.2; CUDA ?, TensorRT ?, cuDNN ?, Ubuntu 22.04. | A |
| module support | PASS | Jetson AGX Orin 32GB is in NVIDIA's supported-module list for JetPack 6.2.3. | A |
| install method | PASS | SDK Manager (host flash): Host-driven flash from an x86 Ubuntu machine (or a Docker container per NVIDIA's SDK Manager Docker image). | A |
| power mode | PASS | 30W (nvpmodel id 2) is an NVIDIA preset for Jetson AGX Orin 32GB on JetPack 6.2.3 (default preset assumed). | A |
| runtime | REQUIRES_VALIDATION | TensorRT version for JetPack 6.2.3 is not listed in the archive. | A |
Verify plan
Run these on the board, in order, to confirm the configuration matches what is checked above.
- cat /etc/nv_tegra_release
Expected:# R36 (release), REVISION: 5.2 …
Settles: jetpack release, install method · community (aggregated forum examples) · forums.developer.nvidia.com/t/379754 · verified 2026-09-07 - dpkg -l nvidia-jetpack (or: apt-cache show nvidia-jetpack; dpkg-query --show nvidia-l4t-core)
Expected:ii nvidia-jetpack <version> arm64 NVIDIA Jetpack Meta Package
Settles: jetpack release · community (Collabnix / NVIDIA forum aggregation) · forums.developer.nvidia.com/t/145422 · verified 2026-09-07 - cat /proc/device-tree/model
Expected:NVIDIA Jetson <board name> Developer Kit (Super-flashed boards append 'Super' or 'Engineering Reference Developer Kit Super')
Settles: module support, install method · community · medium.com/@arnav04verma/bootstrapping-a-headless-jetson-orin-nano-super-for-uav-autonomy- · verified 2026-09-07 - sudo nvpmodel -q
Expected:NV Power Mode: <MODE_NAME> <mode_id>
Settles: power mode · community (JetsonHacks / RidgeRun aggregation) · jetsonhacks.com/2018/10/07/nvpmodel-nvidia-jetson-agx-xavier-developer-kit · verified 2026-09-07 - nvcc --version
Expected:Cuda compilation tools, release <major.minor>, V<full version>
Settles: jetpack release · community (jetsonhacks/jetsonUtilities aggregation) · github:jadwigo/86b905ca2573dc7b9a685652b82ef590 · verified 2026-09-07 - dpkg -l | grep tensorrt
Expected:ii tensorrt <version> arm64 ... (and libnvinfer*, python3-libnvinfer* rows)
Settles: jetpack release, runtime · community (aggregated Jetson verification guides) · github:jadwigo/86b905ca2573dc7b9a685652b82ef590 · verified 2026-09-07 - python3 -c "import tensorrt; print(tensorrt.__version__)"
Expected:<major.minor.patch> (e.g. 10.3.0)
Settles: runtime · community · nvidia-docs:deeplearning/tensorrt/latest/getting-started/release-notes-10/10.13.2.html · verified 2026-09-07
Remediations
No remediation needed — every check passed.
Known issues
No known issues apply to this configuration.
nvpmodel presets on JetPack 6.2.3
| Mode | nvpmodel id | Watts | Flash | |
|---|---|---|---|---|
| MAXN | 0 | — | standard | |
| 15W | 1 | 15 | standard | |
| 30W | 2 | 30 | standard | default |
| 40W | 3 | 40 | standard |
Full detail: Jetson AGX Orin 32GB on the power-modes page.
Other install methods
This page checks the SDK Manager (host flash) path. How the other install methods behave for Jetson AGX Orin 32GB on JetPack 6.2.3:
| Method | Notes |
|---|---|
| SD card image | Not offered for Jetson AGX Orin 32GB. |
| ISO image (target-side) | Available; cannot upgrade a non-Super board into Super Mode — reflash from a host to change flash configuration. |
| apt upgrade / OTA | Works only within Jetson Linux 6.x; crossing a major L4T version needs a full reflash. |
| flash.sh / Jetson Linux BSP | Always available from an x86 host; the only command-line way to select a custom flash configuration (e.g. Super). |
Sources
Sources
- NVIDIA · nvidia-dev:embedded/jetpack-archive · verified 2026-09-07 · source · class A
JetPack 6.2.3 | L4T 36.5.2 — Supports: Jetson AGX Orin Series, Jetson Orin NX Series, Jetson Orin Nano Series
- NVIDIA · nvidia-docs:jetson/r36.4.4/DeveloperGuide · verified 2026-09-07 · source · class A
MAXN_SUPER mode is available only when flashing with jetson-orin-nano-devkit-super or jetson-orin-nano-devkit-super-maxn configuration.
- NVIDIA · jetson-linux-r36.4.4/power-orin r36.4.4 (JetPack 6.2.1) · NVP Model Clock Configuration for Jetson AGX Orin 32GB · verified 2026-09-07 · source · class A
MAXN | 15W | 30W* | 40W — The default mode is 30W (mode ID 2).
- NVIDIA · DS-10662-001 v1.8 · Power, Table · verified 2026-09-07 · source · class A
JAO 32GB Maximum Module Power: Up to 40W
Assumptions
- JetPack 6.2.3 = Jetson Linux 36.5.2; CUDA ?, TensorRT ?, cuDNN ?, Ubuntu 22.04.
- Jetson AGX Orin 32GB is in NVIDIA's supported-module list for JetPack 6.2.3.
- SDK Manager (host flash): Host-driven flash from an x86 Ubuntu machine (or a Docker container per NVIDIA's SDK Manager Docker image).
- 30W (nvpmodel id 2) is an NVIDIA preset for Jetson AGX Orin 32GB on JetPack 6.2.3 (default preset assumed).
- TensorRT version for JetPack 6.2.3 is not listed in the archive.
Warnings
- NVIDIA's archive lists no cuda / tensorrt / cudnn versions for JetPack 6.2.3; read them off the board (nvcc --version, dpkg -l | grep tensorrt).
Method and limitations
Each check compares the requested configuration against NVIDIA's published facts (JetPack archive and release notes, Jetson Linux developer guide, power-mode registry, developer-forum staff replies): PASS when it is a published pairing, VERSION_RISK when it works but is not the officially paired version, REQUIRES_VALIDATION when the published data cannot decide it, UNSUPPORTED when NVIDIA says it does not work, UNKNOWN when the input is not in the registry. The verdict is the worst check. Every check names the on-device commands that settle it, with the expected output.
Full method, status vocabulary, verdict rule and the verify-command table: Jetson Configuration Checker methodology.
Links
- Live engine (this exact configuration): https://edgeaistack.ai/engines/jetson-configuration-checker/?hw=jetson_agx_orin_32gb&jp=6.2.3&install=sdk_manager&rt=tensorrt&mv=1.0&dv=2026-09-07
- Methodology: /methodology/jetson-configuration-checker/
- Dataset: /datasets/jetpack-config.json
- Power modes: /platforms/jetson-power-modes/
- Jetson AGX Orin 32GB overview: /compatibility/jetson-agx-orin-32gb/
Change one input and re-run.
The live engine keeps every source line and gives you a fresh permalink and summary.