Difference between revisions of "Jetson AGX Xavier"

From eLinux.org
Jump to: navigation, search
m
m
Line 1: Line 1:
 
NVIDIA [https://developer.nvidia.com/embedded/buy/jetson-xavier-devkit Jetson AGX Xavier] is an embedded system-on-module (SoM) from the NVIDIA [https://www.nvidia.com/en-us/deep-learning-ai/products/agx-systems/ AGX Systems] family, including an integrated Volta GPU with Tensor Cores, dual Deep Learning Accelerators (DLAs), octal-core NVIDIA Carmel ARMv8.2 CPU, 16GB 256-bit LPDDR4x with 137GB/s of memory bandwidth, and 650Gbps of high-speed I/O including PCIe Gen 4 and 16 camera lanes of MIPI CSI-2.
 
NVIDIA [https://developer.nvidia.com/embedded/buy/jetson-xavier-devkit Jetson AGX Xavier] is an embedded system-on-module (SoM) from the NVIDIA [https://www.nvidia.com/en-us/deep-learning-ai/products/agx-systems/ AGX Systems] family, including an integrated Volta GPU with Tensor Cores, dual Deep Learning Accelerators (DLAs), octal-core NVIDIA Carmel ARMv8.2 CPU, 16GB 256-bit LPDDR4x with 137GB/s of memory bandwidth, and 650Gbps of high-speed I/O including PCIe Gen 4 and 16 camera lanes of MIPI CSI-2.
 +
 +
[[File:Xavier-module-dev-kit-3qrtr-1945px.png|800px|right|text-bottom]]
  
 
Useful for deploying computer vision and deep learning to the edge, Jetson AGX Xavier runs Linux and provides 32 TeraOPS of compute performance in user-configurable 10/15/30W power profiles.
 
Useful for deploying computer vision and deep learning to the edge, Jetson AGX Xavier runs Linux and provides 32 TeraOPS of compute performance in user-configurable 10/15/30W power profiles.
Line 5: Line 7:
 
Jetson AGX Xavier is currently available as the [https://developer.nvidia.com/embedded/buy/jetson-xavier-devkit Jetson AGX Xavier Developer Kit], with the OEM compute module becoming available in the future.  See the wiki of previous Jetson's [[Jetson|here]].
 
Jetson AGX Xavier is currently available as the [https://developer.nvidia.com/embedded/buy/jetson-xavier-devkit Jetson AGX Xavier Developer Kit], with the OEM compute module becoming available in the future.  See the wiki of previous Jetson's [[Jetson|here]].
  
[[File:Xavier-module-dev-kit-3qrtr-1945px.png|800px|right|text-bottom]]
+
= Jetson AGX Xavier Module =
 +
 
 +
The Jetson AGX Xavier compute module contains all the active processing components.  The I/O ports are broken out through a carrier board via a 699-pin board-to-board connector.
 +
 
 +
Below is a partial list of the module's features.  Please see the [http://developer.nvidia.com/embedded/dlc/jetson-xavier-data-sheet Jetson AGX Xavier Module Datasheet] for the complete specifications.
 +
 
 +
[[File:Jetson Xavier Module Block Diagram.jpg|700px|right]]
 +
 
 +
=== Processing Components ===
 +
* Octal-core NVIDIA Carmel ARMv8.2 CPU @ 2.26GHz
 +
* 512-core Volta GPU @ with 64 Tensor Cores
 +
* Dual [http://nvdla.org/ Deep Learning Accelerator] (DLA) engines
 +
* 16GB 256-bit LPDDR4x @ 2133MHz (137GB/s)
 +
* 32GB eMMC 5.1
 +
* Vision Accelerator engine
 +
* (4x) 4Kp60 H.264/H.265 video encoder
 +
* (2x) 8Kp30 / (6x) 4Kp60 H.264/H.265 video encoder
 +
 
 +
[[File:Xavier-module-standing-3qrtr-alpha-1500px.jpg|350px|right]]
 +
 
 +
=== I/O Interfaces & Ports ===
 +
* (16x) MIPI CSI-2 lanes, (8x) SLVS-EC lanes
 +
** up to 6 active sensor streams and 36 virtual camera channels
 +
* (5x) PCIe Gen 4 controllers | 1x8, 1x4, 1x2, 2x1
 +
** (3x) Root Port & Endpoint
 +
** (2x) Root Port
 +
* (3x) USB 3.1 + (4x) USB 2.0
 +
* (3x) eDP 1.4 / DP 1.2 / HDMI 2.0 @ 4Kp60
 +
* 10/100/1000 BASE-T Ethernet + MAC + RGMII PHY
 +
* Dual CAN bus controller
 +
* UART, SPI, I2C, I2C, GPIOs
 +
 
 +
=== Form-Factor ===
 +
* 699-pin board-to-board connector
 +
* dimensions: 100x87mm with 16mm Z-height
 +
* Integrated Thermal Transfer Plate (TTP) with heatpipe
 +
* -25C to 80C operating temperature
 +
* 9.0-20VDC input power
 +
 
 +
=== Software Support ===
 +
<div style="width:40%;column-count:2;-moz-column-count:2;-webkit-column-count:2">
 +
* [https://developer.nvidia.com/embedded/jetpack JetPack 4.0 EA]
 +
* [https://developer.nvidia.com/embedded/linux-tegra Linux4Tegra R31.0.1] (L4T)
 +
* Linux kernel 4.9
 +
* Ubuntu 18.04 aarch64
 +
* [https://developer.nvidia.com/cuda-toolkit CUDA Toolkit] 10.0
 +
* [https://developer.nvidia.com/cudnn cuDNN] v7.3
 +
* [https://developer.nvidia.com/tensorrt TensorRT] 5.0 RC
 +
* [https://developer.nvidia.com/embedded/visionworks VisionWorks] 1.6
 +
* OpenCV 3.3.1
 +
* OpenGL 4.6 / OpenGL ES 3.2
 +
* Vulkan 1.0
 +
* [http://developer.nvidia.com/embedded/dlc/l4t-multimedia-api-reference-24-2-1 L4T Multimedia API] (Argus 0.97)
 +
* GStreamer 1.14.1
 +
* V4L2 media controller support
 +
* [https://developer.nvidia.com/nsight-systems NVIDIA Nsight Systems] 2018
 +
* [https://developer.nvidia.com/nsight-graphics NVIDIA Nsight Graphics] 1.0
 +
</div>
 +
 
 +
= Jetson AGX Xavier Developer Kit =

Revision as of 13:08, 18 September 2018

NVIDIA Jetson AGX Xavier is an embedded system-on-module (SoM) from the NVIDIA AGX Systems family, including an integrated Volta GPU with Tensor Cores, dual Deep Learning Accelerators (DLAs), octal-core NVIDIA Carmel ARMv8.2 CPU, 16GB 256-bit LPDDR4x with 137GB/s of memory bandwidth, and 650Gbps of high-speed I/O including PCIe Gen 4 and 16 camera lanes of MIPI CSI-2.

Xavier-module-dev-kit-3qrtr-1945px.png

Useful for deploying computer vision and deep learning to the edge, Jetson AGX Xavier runs Linux and provides 32 TeraOPS of compute performance in user-configurable 10/15/30W power profiles.

Jetson AGX Xavier is currently available as the Jetson AGX Xavier Developer Kit, with the OEM compute module becoming available in the future. See the wiki of previous Jetson's here.

Jetson AGX Xavier Module

The Jetson AGX Xavier compute module contains all the active processing components. The I/O ports are broken out through a carrier board via a 699-pin board-to-board connector.

Below is a partial list of the module's features. Please see the Jetson AGX Xavier Module Datasheet for the complete specifications.

Jetson Xavier Module Block Diagram.jpg

Processing Components

  • Octal-core NVIDIA Carmel ARMv8.2 CPU @ 2.26GHz
  • 512-core Volta GPU @ with 64 Tensor Cores
  • Dual Deep Learning Accelerator (DLA) engines
  • 16GB 256-bit LPDDR4x @ 2133MHz (137GB/s)
  • 32GB eMMC 5.1
  • Vision Accelerator engine
  • (4x) 4Kp60 H.264/H.265 video encoder
  • (2x) 8Kp30 / (6x) 4Kp60 H.264/H.265 video encoder
Xavier-module-standing-3qrtr-alpha-1500px.jpg

I/O Interfaces & Ports

  • (16x) MIPI CSI-2 lanes, (8x) SLVS-EC lanes
    • up to 6 active sensor streams and 36 virtual camera channels
  • (5x) PCIe Gen 4 controllers | 1x8, 1x4, 1x2, 2x1
    • (3x) Root Port & Endpoint
    • (2x) Root Port
  • (3x) USB 3.1 + (4x) USB 2.0
  • (3x) eDP 1.4 / DP 1.2 / HDMI 2.0 @ 4Kp60
  • 10/100/1000 BASE-T Ethernet + MAC + RGMII PHY
  • Dual CAN bus controller
  • UART, SPI, I2C, I2C, GPIOs

Form-Factor

  • 699-pin board-to-board connector
  • dimensions: 100x87mm with 16mm Z-height
  • Integrated Thermal Transfer Plate (TTP) with heatpipe
  • -25C to 80C operating temperature
  • 9.0-20VDC input power

Software Support

Jetson AGX Xavier Developer Kit