NVIDIA900-82888-0000-000 | Jetson AGX Xavier ModuleSystem-on-Module- SOM

The Jetson AGX Xavier Module is a computational accelerator designed for use in AI applications involving image recognition, object detection, and speech recognition. The Jetson AGX Xavier Module provides 20× the performance and 10× the energy efficiency compared to its predecessor, the Jetson TX2, and consumes under 30 W of power. The Jetson AGX Xavier Module contains a 512-core NVIDIA Volta GPU with Tensor Cores, 2× NVDLA engines, an eight-core Arm V8.2 64-bit Cameral CPU, 16 GB of 256-bit LPDDR4x, 32-GB eMMC 5.1 flash storage, and seven-way VLIW vision-accelerator processor. The Jetson AGX Xavier Module is capable of up to 32 TOPS and features multiple connectivity options, including 8× PCIe Gen 4, USB 3.1, single-lane UFS, UART, SPI, CAN, I2C, I2S, DMIC, and GPIO. The Jetson AGX Xavier Module also integrates a Gigabit Ethernet and has a 699-pin Molex Mirror Mezz Connector for connecting to external boards. The board has an operating temperature range of –25°C to 80°C, can operate at 95% relative humidity, and has a vibration of 5 G RMS and a shock value of 50 G. The Jetson AGX Xavier Module supports a range of frameworks and platforms, including Linux, NIVIDA CUDA, cuDNN, and TensorRT software libraries.

Drawings/Package

      

Key Features

• Up to 32 TOPS

• 512-core NIVIDA Volta GPU

• Eight-core 64-bit Arm CPU

• Maximum 30-W power

• Supports Ubuntu Linux OS, NIVIDIA CUDA, cuDNN, and TensorRT

General Applications

• Object recognition

• Speech recognition

• Automation

• Agriculture

• Deep learning

Evaluation Board

YES — The Jetson AGX Xavier Module can be evaluated with the Jetson AGX Xavier Developer Kit available from Arrow.