Hardware Dev kits

One of the applications I’m interested in working on is one that could be considered an edge application, in that once it is trained it could run on stand-alone hardware. Thus I am considering one of the nVidia Jetson dev kits. I noticed there are three options, the Jetson Nano, the newly released Jetson Xavier NX and lastly the Jetson AGX Xavier.

I’ve seen posts on the Jetson nano, but nothing yet on the Xavier NX (forgive me if I missed one). Because of the ram size of the Xavier NX of 8 gb, vs 4 gb of the Nano I would think that would be the most reasonable size to get.

Now, my particular application has to do with images, so I’m also concerned with the data size. That said I don’t know if I really need the added memory of 32 gb on the AGX Xavier over the NX. Ideally, once the application is trained, I’d like to deploy it on something akin to the Jetson Nano. The application would be doing something similar to superresolution and ‘decrappifcation’ at a refresh rate of four times a second or slower.

Also, if no one knows already, maybe I can TOFTT and test it out.

1 Like

So I ended up getting the Nano for now. The reason is the application I’m working on should only need that level of power for inference, so I’d like to test it.

That said, once I get up and running I might test out the Xavier NX as well to see if it can handle a stretch goal for this application.

So far I have not been able to play with the Jetson, but I did review many tutorials and nVidia’s own presentations. I did note that there is a roadmap where they show what is up-coming. It seems that the Nano Next will be out next year:

I’m hoping it will be using the new Ampere architecture, but that is just a hope, I haven’t seen anything to suggest it will.

Also, about the Xavier NX board, I really like the fact that it has an m.2 slot for storage. It will be really important for my stretch goal to be able to use m.2 storage reliably.

1 Like