Search / Catalog / SOM / CoM / PW-SOM-102
SOM / CoMSupply confirmed per project NVIDIA JETSON · PREVIOUS GENERATION

NVIDIA Jetson AGX Xavier

PW-SOM-102

Official NVIDIA module (previous generation) — NVIDIA Jetson AGX Xavier (NVIDIA Jetson Xavier)

AI performance / GPU
32 TOPS
CPU cores
8-core NVIDIA Carmel Arm v8.2 64
Memory configuration
16GB / 32GB LPDDR4x
Storage
32GB eMMC 5.1
Carrier board for Jetson SO-DIMM modules: Chinchilla Jetson 001(for Orin NX / Orin Nano / Xavier NX / Nano)

NVIDIA Jetson AGX Xavier is a SOM / CoM from NVIDIA. SoC: NVIDIA Jetson AGX Xavier. Form factor: 100 x 87 mm AGX module (699-pin) (100 x 87 mm (board) / 105 x 105 mm (including heat spreader)).

AI performance is 32 TOPS (INT8). Available configurations: 32GB 256bit LPDDR4 x 136.5GB/s memory and 32GB eMMC 5.1 storage. Supported OS: Linux, Ubuntu, JetPack, and L4T.

Intended applications include edge AI, vision, automotive, and smart factory / logistics.

Positive ONE provides technical support. We also handle carrier board design and contract development.

Inquiries about this product

We welcome inquiries about technical specifications, pricing, lead time, and evaluation unit loans. We also handle carrier board design and contract development.

Product overview

The NVIDIA Jetson AGX Xavier module enables AI-powered autonomous machines. It runs on as little as 10W and delivers up to 32 TOPS of AI performance. As a world-leading AI computing platform, NVIDIA’s rich AI tools and workflows let developers quickly train and deploy neural networks. It supports the NVIDIA JetPack SDK, reducing development effort and cost.

The Jetson AGX Xavier series delivers cutting-edge compute density, power efficiency, and AI inference performance. It provides preset (configurable) power profiles of 10W/15W/30W, and Jetson AGX Xavier Industrial supports 20W/40W profiles. These can be switched at runtime and customized to your requirements. Compared to Jetson TX2, it achieves 10x the energy efficiency and 20x the performance.

Key features

Video encoding

  • 2 × 1000 MP/s
  • 4K60: 4x (HEVC)
  • 4K30: 8x (HEVC)
  • 1080p60: 16x (HEVC)
  • 1080p30: 32x (HEVC)

Video decoding

  • 2 × 1500 MP/s
  • 8K30: 2x (HEVC)
  • 4K60: 6x (HEVC)
  • 4K30: 12x (HEVC)
  • 1080p60: 26x (HEVC)
  • 1080p30: 52x (HEVC)
  • 1080p30: 30x (H.264)

Included items

  • 1 × NVIDIA Jetson AGX Xavier module

Configuration stage (implementation stage)

SOM implementation progresses in the order “module alone → mounted on a carrier board → enclosed”. The position of this product is shown below.

STEP 1
SOM only
System on Module
A core module that integrates the SoC, memory, and power supply. Used mounted on a carrier board.
This product is here
→
STEP 2
SOM + carrier board
Carrier / SBC / Evaluation board
A working board that carries the SOM and brings out external I/O (Ethernet, USB, video, power, etc.). For evaluation and integration.
→
STEP 3
SOM + carrier board + enclosure
Box PC / AI BOX / Edge server
Finished units with the board housed in an enclosure, designed for heat dissipation, dust protection, and installation. They can be installed and put into mass production as-is.
This product is STEP 1 (SOM only). It can be mounted on a carrier board for STEP 2 and housed in an enclosure for STEP 3. Positive ONE can support you end to end, including carrier board design and enclosure integration (contract development).

SILICON / SoC specifications

ModuleNVIDIA Jetson AGX Xavier
SoC seriesNVIDIA Jetson AGX
CPU8-core NVIDIA Carmel Arm v8.2 64-bit
GPU512-core Volta + 64 Tensor Core
AI performanceUp to 32 TOPS
DLA2x NVDLA
RAM configuration16GB / 32GB LPDDR4x
Storage32GB eMMC 5.1
OS supportLinux / Ubuntu / JetPack / L4T
Form factor100 x 87 mm AGX module (699-pin)
Outline100 × 87 mm
AvailabilityPrevious generation; availability confirmed per long-term project

Jetson AGX Xavier series specifications

Jetson AGX Xavier series specifications
AGX Xavier IndustrialAGX Xavier (64GB)AGX Xavier (32GB)
AI performance30 TOPS32 TOPS32 TOPS
GPU512-core NVIDIA Volta GPU + 64 Tensor cores
Max GPU frequency1211 MHz1377 MHz1377 MHz
CPU8-core NVIDIA Carmel Arm v8.2 64-bit / 8MB L2 + 4MB L3
Max CPU frequency2.0 GHz2.2 GHz2.2 GHz
DL accelerator2x NVDLA
Max DLA frequency1.2 GHz1.4 GHz1.4 GHz
Vision accelerator2x PVA
Safety cluster engine2x Arm Cortex-R5 lockstep--
Memory32GB 256bit LPDDR4x (ECC) 136.5GB/s64GB 256bit LPDDR4x 136.5GB/s32GB 256bit LPDDR4x 136.5GB/s
Storage64GB eMMC 5.132GB eMMC 5.132GB eMMC 5.1
Video encoding2x 4K60 | 6x 4K30 | 12x 1080p60 | 24x 1080p30 (H.265)4x 4K60 | 8x 4K30 | 16x 1080p60 | 32x 1080p30 (H.265)—
Video decoding2x 8K30 | 4x 4K60 | 8x 4K30 | 18x 1080p60 | 36x 1080p30 (H.265)2x 8K30 | 6x 4K60 | 12x 4K30 | 26x 1080p60 | 52x 1080p30 (H.265)—
CSI cameraUp to 6 cameras (36 virtual) 16-lane MIPI CSI-2 D-PHY 1.2(40Gbps) C-PHY 1.1(62Gbps)Up to 6 cameras (36 virtual) 16-lane MIPI CSI-2 | 8-lane SLVS-EC—
PCIe1x8 + 1x4 + 1x2 + 2x1 (PCIe Gen4, Root Port & Endpoint)
USB3x USB 3.2 Gen2 (10Gbps) / 4x USB 2.0
Networking1x GbE
Display3x multi-mode DP 1.4 / eDP 1.4 / HDMI 2.0
Other I/O5x UART, 3x SPI, 4x I2S, 8x I2C, 2x CAN, PWM, DMIC, GPIO
Power consumption20W - 40W10W - 30W10W - 30W
Outline / connectors100mm × 87mm / 699-pin connector / integrated thermal transfer plate

* NVIDIA official specifications. 100mm×87mm, 699-pin connector.

Characteristics / application / solutions

32-TOPSLarge AGXVolta-GPUCarmel-CPU10-30W699-pinPrevious generation
Main applications
Legacy design maintenanceAutonomous machinesRoboticsEdge inference
Supported solutions
AI VisionRobotics & AMREdge Server

Same series / related products