HiJetsonHacks! Can you do a video on how to start and run a simple program from the host machine (ubuntu 18.04) using Nsight and execute it on the Jetson Xavier. Having issues achieving the simplest of things on the Jetson Xavier. Too many setting to jump through.
Thank you for the video. Following your footsteps... :-) How do we run the demos/samples? Although I selected to install the whole JetPack, I can't find them. Also, is it possible to enable audio via HDMI on Xavier? It's listed in the sound devices but it doesn't work.
Hi, thank you for this amazing video. It really helps a lot. But I have one basic question. Do we need the Host Ubuntu Machine only once to set everything up on the Jetson Xavier or we need it all the time to do anything on the Jetson? Thanks a lot for your time.
Once you have L4T installed (along with the libraries you have selected during the JetPack installation), you do not need to use the host computer anymore. The Xavier is a full Linux desktop. Thanks for watching!
Thanks Jim for the cool demo!! My installation seems to be Ok, but some samples are missing. Just got nvx_sample_sfm in /home/nvidia/VisionWorks-SFM-0.90-Samples/bin/aarch64/linux/release/ Any hints?
Training a neural network is possible on the Xavier, but remember that there is a limited amount of memory and GPUs in comparison to a normal desktop /laptop with a NVIDIA graphics card. Typically people train in the cloud/desktop/laptop and then deploy inferencing on a Jetson. The Jetson run Ubuntu, so you can run "normal" Python scripts for that environment. Thanks for watching!
@@JetsonHacks Hi, thanks for getting back to me. What about Orion? Should it perform better than a normal PC? Is it cheaper than a laptop with same specs?
@@joao333hd6 Depends on what your definition is for a 'normal' PC. If you do not have a requirement for low power usage (a 32GB Orin uses a maximum of 40W), a normal PC with a NVIDIA 30xx series graphics card should be much faster and less expensive. An Orin is a $2K developer kit. This is as you expect, considering a 3070 GPU card alone uses 220W and has ~ 3X the number of GPU cores. The Jetsons are designed mostly for places where power consumption is important, such as robotics, automotive, and IoT types of applications. That's not to say that you can't train or run complicated programs on the Jetsons (memory on the smaller Jetsons is an issue), just that the cross over point on performance between PCs with discreet GPUs hasn't been reached yet. A higher end Jetson isn't really a PC replacement, it's a different type of animal all together.
Hi again. Thank you very much. I was under the impression that the PC would me much more expensive. Graphics, cpu and etc. Guess I need to check a good build around the 30xx. Thanks, João
Hi Jim, First of all thanks for this interesting video, However after flashing the AGX Xavier and start to run it; it gives me an update message to upgrade the linux 18.04 to the 20.04! should I go for it and accept the upgrade on the AGX Xavier or ignore it since I really dont know what will happen if I do that. Thanks
You can, but you would probably be happier getting a similarly priced notebook because the notebook will have much better support for the tasks you mention. Thanks for watching!
hello jim, i have a AGX Xavier running Ubuntu 20.04 and Jetpack 5, but i wish to upgrade to Ubuntu 22.04 for installing ROS2 Humble.... but i dont know how to do this, and i dont want to mess up this system as it is my work system, can you please guide me how to upgrade to Ubuntu 22.04
AGX Xavier does not have an update path to Ubuntu 22.04. You can use a Docker container to run ROS2 Humble on 20.04: github.com/dusty-nv/jetson-containers/tree/master/packages/ros
nvidia-smi does not run in the Jetson environment. Please use tegrastats instead. Missing nvcc indicates that you have an incomplete installation. Did you install CUDA? On a JetPack 4.1 installation, nvcc should be in /usr/local/cuda-10.0/bin/. Thanks for watching!
Do I need to install drivers to enable a WiFi dongle on the Xavier? I don't have available Ethernet during jetpack install so I flashed and now I'm manually setting up the internet so I can install the other stuff but it doesn't seem to run the network manager, any thoughts?
My understanding is that it depends on your WiFi dongle as to whether or not it needs drivers or not. I have only installed the software on the Jetsons using the recommended procedure, I don't have any experience with not using a hard wired Ethernet connection. Please ask this question on the official NVIDIA Jetson forum where a large number of developers and NVIDIA engineers share their experience. The forum is here: devtalk.nvidia.com/default/board/326/jetson-agx-xavier/
@@JetsonHacks : I followed the hard-wired ethernet connection as documented but it could never identify the IP address or get a response from it on my unit even after entering it manually (after it fails it prompts you for this info). Instead, I gave it the IP address for the wireless M.2 that I had installed (Intel 8265). That allowed the flash to complete. Once booted on the Xavier I can see it knows the hard wired ethernet is there, even lists its IP address. I did *not* need to explicitly install any drivers for that WIFI connection, I assume they were installed by default during the installation.
I picked one of these up for £5 at a car boot from a house clearance. Can I mine dogecoin on it lmao. Or would it make a great media centre in comparison to nvidia shield
It's a developers kit, most things are possible! However, it is for developers, so there's a lot of development skills you will need for projects. Thanks for watching!
3 questions: 1) When you run xavier on first boot ita ask to run installation.sh script, why libraries are not installed with this script? What's the point? 2) Why you can't just add private deb repository into apt repo list and install everything by regular apt-get? This will be A LOT more handy. 3) Can we install proper libraries WITHOUT jetpack? Jetpack is really frustrating solution. I really feel PAIN while watching this video (
These questions are more than can be answered here. Please ask this question on the official NVIDIA Jetson forum where a large number of developers and NVIDIA engineers share their experience. The forum is here: devtalk.nvidia.com/default/board/326/jetson-agx-xavier/
@@JetsonHacks That folder you use /usr/aarch64/linux/release does not exist. I just flashed my jetson tx2 with jetpack 4.2 I was finally able (after a long painful search) to find the make files for the oceanFFT thats fine but what I'm looking for is to run the demo for the people walking and I cant find it. The only VisionWorks demo i see is the cars parked in a parking lot
The video is for JetPack 4.1. I believe the VisionWorks demos have moved. You can look in /usr/share/visionworks. In the source directory there is an install file named 'install-samples.sh'. Run the file with a directory name. Then go to the directory and compile the samples. You end up with the demo nvx_demo_motion_estimation in the bin directory.
@@JetsonHacks I found it and managed to compile and run! I think i know the confusion. On Jetpack 4.2 in the /home directory there is a folder called VisionWorks sFM sample and its just one sample so whenever I was looking for visionworks i thought that was the only sample. Now I know where the samples are located. Thanks!
No, JetPack 4 is for the Xavier only. Later this year NVIDIA is planning to ship a version of JetPack which can setup both the TX2 and Xavier. Thanks for watching!
