Jetson Xavier NX
Introduction
The main goal of this wiki is to provide a summary of the main features available in the NVIDIA Jetson Xavier NX platform as well as its differences with its predecessor, the Jetson AGX Xavier platform.
The Jetson NVIDIA Xavier NX
NVIDIA consider this platform as the world's smallest AI supercomputer for embedded and edge systems, designed in a small form factor system on module (SOM) smaller than a credit card. This platform is capable to run modern neural networks in parallel and to process data from multiple high-resolution sensors, supporting the most popular AI frameworks. Jetson Xavier NX represents an excelent option for high-performance AI systems like drones, portable medical devices, small commercial robots, smart cameras, high-resolution sensors, automated optical inspection, and other IoT embedded systems.
Key Features
NVIDIA designed this platform emphasizing in the following three key features:
Size
The NVIDIA Jetson Xavier NX has a 70 mm x 45 mm SoC where the power of an NVIDIA Xavier SoC is packed into a module the size of a Jetson Nano. This small module combines exceptional performance and power advantages with a rich set of IOs—from high-speed CSI and PCIe to low-speed I2Cs and GPIOs. We are now able to take advantage of the small form factor, sensor-rich interfaces, and big performance to bring new capability to all our embedded AI and edge systems.
Performance
This platform delivers 14 TOPS at 10 W and 21 TOPS at 15 W, making it ideal for size and power-constrained systems. With 384 CUDA Cores, 48 Tensor Cores, and 2 NVDLA engines, it can run multiple modern neural networks in parallel and process high-resolution data from multiple sensors simultaneously.
Power
The Jetson Xavier NX platform enable the entire NVIDIA software stack that runs all modern AI networks and frameworks with accelerated libraries for deep learning as well as computer vision, computer graphics, multimedia, and more. And it still leaves more of the available power budget for the sensors and peripherals required by today's popular embedded applications.
Platform Technical Specs
| 10 W | 15 W | |
|---|---|---|
| AI Performance | 14 TOPS (INT8) | 21 TOPS (INT8) |
| GPU | 384-core NVIDIA Volta™ GPU with 48 Tensor Cores | |
| GPU Max Freq | 800 MHz | 1100 MHz |
| CPU | 6-core NVIDIA Carmel ARM®v8.2 64-bit CPU 6MB L2 + 4MB L3 | |
| CPU Max Freq | 2-core @ 1500MHz 4-core @ 1200MHz |
2-core @ 1900MHz 4/6-core @ 1400Mhz |
| Memory | 8 GB 128-bit LPDDR4x @ 1600 MHz 51.2GB/s | |
| Storage | 16 GB eMMC 5.1 | |
| Power | 10W/15W | |
| PCIe | 1 x1 + 1x4 (PCIe Gen3, Root Port & Endpoint) | |
| CSI Camera | Up to 6 cameras (36 via virtual channels) 12 lanes MIPI CSI-2 D-PHY 1.2 (up to 30 Gbps) | |
| Video Encode | 2x464MP/sec (HEVC) 2x 4K @ 30 (HEVC) 6x 1080p @ 60 (HEVC) 14x 1080p @ 30 (HEVC) | |
| Video Decode | 2x690MP/sec (HEVC) 2x 4K @ 60 (HEVC) 4x 4K @ 30 (HEVC) 12x 1080p @ 60 (HEVC) 32x 1080p @ 30 (HEVC) 16x 1080p @ 30 (H.264) | |
| Display | 2 multi-mode DP 1.4/eDP 1.4/HDMI 2.0 | |
| DL Accelerator | 2x NVDLA Engines | |
| Vision Accelerator | 7-Way VLIW Vision Processor | |
| Networking | 10/100/1000 BASE-T Ethernet | |
| Mechanical | 45 mm x 69.6 mm 260-pin SO-DIMM connector | |
Interfaces
| Category | Function | Category | Function |
|---|---|---|---|
| USB | USB 2.0 interface (3x) USB 3.1 (1x) |
UART | 3x |
| PCIe | PCIe (x1 and x4) | SPI | 2x |
| Camera | CSI (3 x4 or 6 x2) Control, clock |
CAN | 1x |
| Display | HDMI/eDP/DP (2x) DP_AUX/HPD (2x), CEC (1x) |
Wi-Fi/BT/Modem | PCIe/UART/I2S, control/handshake |
| Audio | I2S interface (2x) and clock Master clock |
Fan | FAN PWM and tach input |
| SD Card/SDIO | SD Card or SDIO interface (1x) | Debug | UART |
| LAN | Gigabit ethernet | System | Power control, reset, alerts |
| I2C | 4x | Power | Main input and battery back-up for RTC |
Block Diagram
The hardware described in the key features and interfaces table can be observed in the following block diagram:
Jetson Family Comparison
| Jetson Nano | Xavier NX (10 W) | Xavier NX (15 W) | AGX Xavier | |
|---|---|---|---|---|
| CPU | 4x Cortex-A57 @ 1.43GHz | 4x/ Carmel @ 1.2GHz or 2x Carmel @ 1.5GHz |
4x/6x Carmel @ 1.4GHz or 2x Carmel @ 1.9GHz |
8x Carmel @ 2.26GHz |
| GPU | Maxwell, 128 Cores @ 920MHz | Volta, 384 Cores @ 800MHz | Volta, 384 Cores @ 1100MHz | Volta, 512 Cores @ 1377MHz |
| Accelerators | N/A | 2x NVDLA | 2x NVDLA | |
| Memory | 4GB LPDDR4, 64-bit bus (25.6 GB/sec) |
8GB LPDDR4X, 128-bit bus (51.2 GB/sec) |
16GB LPDDR4X, 256-bit bus (137 GB/sec) | |
| Storage | 16GB eMMC | 8GB eMMC | 32GB eMMC | |
| AI Performance | N/A | 14 TOPS | 21 TOPS | 32 TOPS |
| Dimensions | 45mm x 70mm | 45mm x 70mm | 100mm x 87mm | |
| TDP | 10W | 10W | 15W | 30W |
| Module Price | $149 | $499 | $999 | |
| DevKit Price | $99 | $699 | ||
Compatibility with Jetson AGX Xavier
An alternative to start developing for Jetson Xavier NX is using the Jetson AGX Xavier Developer Kit. You just need to download this software patch and install it on your Jetson Xavier AGX Developer Kit.
Where to Buy the Jetson Xavier NX
The Jetson Xavier NX Module has a cost of $459 and can be purchased from the Arrow page.
RidgeRun Support
RidgeRun provides support for embedded Linux development for NVIDIA's platforms, including the new Jetson Xavier NX. We are specialized in the use of hardware accelerators in multimedia applications. RidgeRun's products take full advantage of the accelerators that NVIDIA exposes to perform transformations on the video streams achieving great performance on complex processes.
Contact Us
RidgeRun is an official NVidia Partner and we have created this documentation to support our joint customers. If you have any questions about the content, please let us know through our contact us page.
