Nvidia Jetson Module to/from communication with Industrial HMI in a PLC

[sale2]

Hello,
I want to accumulate data from a HMI of a PLC on cloud or a on-premise workstation. HMI have ethernet option and I can remotely access the HMI for remote access through port forwarding (TCP/IP).
I am looking for a suitable Jetson device. As the HMI have ethernet port, I was wondering if Nvidia Jetson TX2 Module can serve as a gateway device to send data from the HMI to a cloud, and then send the predictions from the model running in the cloud to the HMI.

Has anyone tried this yet?

[linuxdev]

I have not tried. All of the Jetsons have at least gigabit wired ethernet. I’d suggest though that you go to Orin (or at least Xavier) despite the price difference. The software and update for TX2 and older is on maintenance and no feature updates exist. Xavier is currently actively developed, but the next major release will probably not update Xavier (or after that release, not sure). Orin has a Nano which is far more powerful than a TX2 and is actively developed (the Orin Nano is the least expensive of the Orin series).

The part which is hard to answer is if HMI requires anything else other than networking. Certainly the networking is there, but if you need particular software, then you should check that out ahead of time. For example, Jetsons use a 64-bit ARM architecture, and perhaps you have binary-only software that won’t run on that architecture (don’t know). Also, the GPU in Jetsons are integrated directly to the memory controller (iGPU), whereas a desktop GPU is discrete (a dGPU is plugged into a PCI bus). Detection of the GPU sometimes uses nvidia-smi, which in turn depends on a dGPU (software requiring this might not be able to find the GPU).

Another thing to consider is that different versions of CUDA are available on different platforms. TX2 only has CUDA 10, whereas Xavier and Orin can run CUDA 11. Xavier and Orin can make CUDA 12 available via docker. Xavier will never go beyond CUDA 12, but it is likely that as a new major release of CUDA comes out Orin will get that. If you are writing your own software, or using someone else’s, you might need to be careful about release versions.