The $399 NVIDIA Jetson Xavier NX Developer Kit is the new Jetson powerhouse on the block. Looky here:
The Jetson Xavier NX dev kit brings Jetson Xavier performance to help solve AI and robotics where you need some serious machine learning horsepower.
The entry level Jetson Nano is a good way to start for a lot of people, introducing the fundamentals of machine learning and GPU computing power. Consider the Xavier NX as a professional level to that, where you need beast mode to get serious work done.
Like other Jetsons, the Xavier NX dev kit can be thought of in two parts. The first part, the Jetson NX Module contains the compute and memory components. The second part is the carrier board, which provides affordance for connecting peripherals and providing power input.
The Jetson Xavier NX carrier board is the same layout as the Jetson Nano, with a couple of nice changes. First, there is a wireless card in the M.2 Key E slot on the underside of the board, pre-installed with antennas captured by the plastic base. Second, there is a M.2 Key M slot, also on the other side of the carrier board, which affords the means to install expansion items such as a NVMe SSD.
Like the Jetson Nano, the Jetson Xavier NX runs from a micro SD card. Unlike the Nano, the Xavier NX only runs from the supplied 19V power supply through the barrel jack. While you can power the Nano via the micro USB 2.0 port, the Xavier needs more power than can be supplied over that port. The Xavier runs in either 10W or 15W power profiles, the carrier board supports up to 5A@19V.
Here’s a list of items we used in the video:
- NVIDIA Jetson Xavier NX Developer Kit
- Western Digital NVMe SSD
- A faster SSD: Samsung NVMe SSD
- Samsung EVO 64 micro SD card
- Dymo Label Maker
The Jetson Xavier NX includes special purpose machine learning hardware, including 48 Tensor Cores and 2 NVIDIA Deep Learning Accelerators Engines (NVDLA). Overall performance in machine learning tasks averages over 10x that of a Jetson TX2.
The folks over at NVIDIA wrote a great article “Bringing Cloud-Native Agility to Edge AI Devices with the NVIDIA Jetson Xavier NX Developer Kit” that goes over a lot of specs and performance information. Well worth the read.
Here’s a short version of the specs:
|GPU||NVIDIA Volta architecture with 384 NVIDIA CUDA® cores and 48 Tensor cores|
|CPU||6-core NVIDIA Carmel ARM®v8.2 64-bit CPU 6 MB L2 + 4 MB L3|
|DL Accelerator||2x NVDLA Engines|
|Vision Accelerator||7-Way VLIW Vision Processor|
|Memory||8 GB 128-bit LPDDR4x @ 51.2GB/s|
|Storage||microSD (not included)|
|Video Encode||2x 4K @ 30 | 6x 1080p @ 60 | 14x 1080p @ 30 (H.265/H.264)|
|Video Decode||2x 4K @ 60 | 4x 4K @ 30 | 12x 1080p @ 60 | 32x 1080p @ 30 (H.265) 2x 4K @ 30 | 6x 1080p @ 60 | 16x 1080p @ 30 (H.264)|
|Camera||2x MIPI CSI-2 DPHY lanes|
|Connectivity||Gigabit Ethernet, M.2 Key E (WiFi/BT included), M.2 Key M (NVMe)|
|Display||HDMI and display port|
|USB||4x USB 3.1, USB 2.0 Micro-B|
|Others||GPIO, I2C, I2S, SPI, UART|
|Mechanical||103 mm x 90.5 mm x 34.66 mm|
Some Pics natch – Click to Expand
The next big push in the Jetson ecosystem is Docker based containers. While Docker support has been on the Jetsons for a few releases, they are now going mainstream. NVIDIA has built a server ecosystem, NVIDIA NGC, which contains pre-trained AI models and other resources which serve as building blocks in AI application development.
A great example of this is shown in the video as a demo. The 4 applications that are running are containers, running 7 machine learning models in total.
This shows the power of the Jetson Xavier architecture, where you get desktop level performance in a power budget of only 15W.
NVIDA has made the demos publicly available. The scripts for the demo are on Github on the NVIDIA-AI-IOT account in the jetson-cloudnative-demo repository.
The NVIDIA Jetson Xavier NX Developer Kit is the real deal for edge applications and robotics. There’s enough horsepower to run several models at once, while at the same time maintaining a very small power budget.
To be clear, this is pro level. If you are just getting started, the Jetson Nano is a good starting place. On the other hand, if you have outgrown the Nano, have experience with machine learning and/or have demanding inferencing applications, certainly checkout the Jetson Xavier NX Developer Kit.
I would love to see a Geekbench test of the Xavier series running Ubuntu. BTW great work!
Thank you for the kind words, and thanks for reading!
Can you run one intance? Just for reference, it would be highly appreciated.
I do not know what this means.
I’m trying to understand whether both an SD and SSD are necessary and what some of the trade offs might be.
Can anything useful be done with just an SD? If an SSD is added, can one go with no, or a smaller SD?
What would you recommend for an initial experimental setup?
Depends on what you are trying to accomplish. You do not need a SSD. However, a NVMe SSD is 5-10x faster, so you get a performance boost. Also, SSDs tend to be available in much larger sizes (a SSD 256GB is smallish, on the consumer size they range up to ~ 2TB). If you go with a SSD, you will still need a SD card to boot the system. Thanks for reading!
Thank you for the kind words, and thanks for reading!
I’ve gotten to running some rviz simulations using the new NX and it’s a new world coming from the jetson nano. I am sure an SSD would help loading, but I think this kind of tech pushes the bottle neck of progress back to software development. Thanks Nvidia!
Great to hear! Thanks for reading!
Please add the PINOUTS for Xavier NX on your site as you have done for the other boards. Your tutorials and scripts are very helpful. Thanks!
I am new to containers – could you explain how I can see the actual code being used for some of these demos so that I can learn from them?
I do not have anything to share in this area. Please ask this question on the official NVIDIA Jetson Xavier NX forum, where a large group of developers and NVIDIA engineers share their experience: https://forums.developer.nvidia.com/c/agx-autonomous-machines/jetson-embedded-systems/jetson-xavier-nx/258
Is the version of openCV installed in the Xavier NX image optimized for using the GPU?
No. Thanks for reading!
Thanks for the info! So still follow the same instructions you gave for using openCV on the nano?
You would probably be better off using the mdegans repo: https://github.com/mdegans/nano_build_opencv
It works for the Xaviers too.
thanks, that is the one I used. jtop now reads CUDA:YES. I haven’t noticed any significant improvement on the performance yet.
You would have to be using CUDA enhanced algorithms in OpenCV to notice any difference.
First of all thank you for an amazing website, I learned so much from your content. I would like to get your take on a project i am working on with the Xavier NX Dev kit. I worked extensively with the Nano but my area is more software development, I am upgrading one of my products to use the Xavier but i am having trouble with powering it with POE, is there any way i can achieve this? if not, then would a POE splitter with output 19v 3.15A be enough? I would appreciate any help. thank you so much!
Could you do some a performance comparison between the various jetson products? I am pretty new to jetson products and am a bit bewildered by how they perform different tasks. Arm cores, Cuda cores, and tensor cores are overwhelming.
FWIW, I am currently playing with a jetson nano on a turtlebot3 chassis and motor controller.
Would it be possible to build a racecar/j or f1tenth with a jetson nano and then upgrade to a jetson xavier-nx as I get a better understanding of performance requirements?
Depends on what task you are doing. Here’s a comparison in some deep learning benchmarks: https://developer.nvidia.com/blog/jetson-xavier-nx-the-worlds-smallest-ai-supercomputer/
I have settled on a plan of attack.
I’ll continue my initial exploration with the jetson nano on the turlebot3. I ordered a realsense camera to replace/augment the LDS-01 that came with the TB3.
Then I’ll start building the F1Tenth/RacecarJ/Racecar and use a jetson nx.
To be honest the jetson nano is not limiting my learning… just my buggies ability to go fast. The only thing better then lots of GB’s is lots of MPH’s.
Sounds like a good plan!
Great site and channel. I even appreciate the mako plush and fire extinguisher ‘dad jokes.’
I hope you can make the Racecar/J site go. One thought might be to develop a line of F2Tenth vehicles…. Kind of like F1 on a tighter budget for people just getting started.
Thank you for the kind words. Right now the supply chains are so backed up, it’s hard to tell when most of the items for the RACECARs will be available. However, a good way to get started is with the DIYRobocars or JetRacer where you can get started for just a few hundred dollars. Thanks for reading!
I had not come across DIYRobocars. Looks like a lot of useful information.
For some reason, I am fascinated about having the vehicle create a map of the track, create an optimized raceline, and then deal with other cars on the track.
Can you imagine how fast an real F1 Car could go if the designers didn’t have to worry about a human driver taking up all that space or delays in human reaction times.
More importantly the tv broadcasters would have to talk about the engineers and software developers instead of rambling on about the drivers 🙂
Can you run an instance of Geekbench  which is a very popular software for benchmarking hardware with various tests. I really wanted to know the results, to check how well it perfoms let’s say against an iphone. 
No thanks. I don’t have a reason to run benchmarks.
Hi ! Your videos are really resourceful ! However, I’m in need of some suggestions. I’m building a ROS based rover, and I can’t decide which main board should I buy ! I’m torn between Xavier NX or an Intel NUC developer kit. The stereo camera i’m going to use is realsense d435i. I have seen some compatibility issues with xavier NX and this camera, on the internet. Can you tell me if all the functionalities of this camera work well with xavier nx? It will be of great help ! Thanks !
You should use whichever you feel most comfortable with. I have not found issues with the D435i on the Jetsons, however I am more familiar with the platform than most. Good luck on your project!
Can you use the production module with the dev kit carrier board?
I believe so, but it is worth checking on the official NVIDIA Jetson Xavier NX forum, where a large group of developers and NVIDIA engineers share their experience: https://forums.developer.nvidia.com/c/agx-autonomous-machines/jetson-embedded-systems/jetson-xavier-nx/258