Jetson Xavier™ NX Adds to Jetson Family
NVIDIA today announced Jetson Xavier™ NX, the world's smallest and most advanced embedded AI supercomputer for autonomous robots and edge computing devices. Jetson Xavier NX delivers server-class performance, up to 21 TOPS at 15W or up to 14 TOPS at 10W in a compact 70x45mm form factor.
Also, if you have the Jetson AGX Xavier developer kit, you can check the specifications of Jetson Xavier NX by applying the emulation package below.
*Move to NVIDIA developer site.
Jetson Xavier NX Overview
Embedded designers can refer to the design collateral for production design and system development of Jetson Xavier NX modules. Pin compatibility with Jetson Nano™ enables upgrade to Jetson Xavier NX.
Software developers can also jump-start building AI applications by using the Jetson AGX Xavier Developer Kit and applying a device configuration patch to JetPack to make the device behave as a Jetson Xavier NX. Through software, you can change the number of CPU and GPU cores available, in addition to setting the core clock frequency and voltage for the entire system. This patch is fully reversible and can be used to approximate Jetson Xavier NX performance prior to hardware availability.
Jetson Xavier NX Specifications
10W MODE |
15W MODE |
|
AI Performance |
14 TOPs (INT8) |
21 TOPs (INT8) |
GPUs |
384-core NVIDIA Volta™ GPU with 48 Tensor Cores |
|
Max GPU Freq |
800MHz |
1100MHz |
CPU |
Up to 6-core NVIDIA Carmel ARM® v8.2 64-bit CPU |
|
CPU Max Freq |
4 core @ 1.2GHz 2 cores @ 1.5GHz |
4/6 core @ 1.4GHz 2 cores @ 1.9GHz |
Memory |
8GB 128-bit LPDDR4x @1600MHz, 51.2GB/s |
|
Storage |
16GB eMMC |
|
PCIe |
1 x1 + 1x4 PCIe, total 80GT/s (Gen3, Root Port & Endpoint) |
|
CSI Camera |
Up to 6 cameras (36 via virtual channels) 12 lanes (3x4 or 6x2) MIPI CSI-2, D-PHY 1.2 (up to 30Gbps) |
|
CSI+VI max throughput |
1.9GPix/s (D-PHY, RAW16) |
|
Video Encode |
2x 4K @ 30 | 6x 1080p @ 60 | 14x 1080p @ 30 (HEVC) |
|
Video Decode |
2x 4K @ 60 | 12x 1080p @ 60 | 32x 1080p @ 30 (HEVC) | 16x 1080p @ 30 (H.264) |
|
Display |
HDMI 2.0 (4Kp60) or DP1.4* (4Kp120) | eDP 1.4 (4Kp120) (2 simultaneous) |
|
DL Accelerator |
2x NVDLA Engines |
|
Networking |
10/100/1000BASE-T Ethernet |
|
USB |
1 USB 3.1 (10Gbps) | 3xUSB 2.0 |
|
Other IOs |
1x SDIO / 2x SPI / 3x UART / 2x I2S / 4x I2C / 1x CAN / GPIOs |
|
Mechanical |
45mm x 69.6mm 260 pin SO-DIMM connector, No TTP |
|
Power |
Up to 10W |
Up to 15W |
Deep Learning Inferencing Benchmarks
Jetson Xavier NX delivers up to 15x higher performance in 25% smaller area at the same power than Jetson TX2. Jetson Nano and Jetson TX2 run on FP16, while Jetson Xavier NX and AGX Xavier run both NVDLA engine and GPU on INT8 simultaneously. The Volta architecture GPU with Jetson Xavier NX's Tensor Cores is capable of computing up to 12.3 TOPS while the module's DLA engines generate up to 4.5 TOPS each.
Release date
Scheduled to start providing in March 2020
price
$399
* This price is the price at the time of shipment of 1K+ (1,000 units or more).
*Please contact us for the price in Japanese Yen.
*All registered trademarks and trademarks belong to their respective owners.
Related links (NVIDIA website)
News release (English)
Jetson Xavier NX Documentation and Design Materials