NVIDIA Jetson Nano: A Feature-Packed Arm Developer Kit For $99 USD

[torsionbar28]

I wonder if Nvidia realizes just how much their logo looks like a snail. They'd probably say nautilus, but I'd say snail.

Visualization of the development pace of their open source drivers

[dungeon]

Yeah. I'd be happy with their crappiest x86 APU slapped together with whatever is laying around since I know they're not really in this market. Also seems like a decent way for them to burn off their remaining FM2 APU stock.

Here you go, for example:

https://www.biostar.com.tw/event/A10N-8800E/

15W FP4 APU or so called mobile as low powered desktop Mobo plus 2x4GB memory would cost you as this nV SOM, both old but new but still would beat it on anything for desktop scenario

There is FP5 with 12W to 54W Ryzens also, but for laptops and embedded V1000 serie and not cheap

I smell like AMD will do something cheap FP5-like for desktop masses this year, so that people could say good bye to Windows 7 more easely

Or maybe pushing with just cheapest excavators for AM4 would be enough

https://www.reddit.com/r/Amd/comments/b1enqq/new_amd_cpu_listed_a69400_bristol_ridge_expected/

[phoron]

Sorry, I lost track... So how many blobs are we talking here ? Would something work with linux-libre and some FSF-compliant distro on this SOC? I vaguely remember someone claiming they had 100% free software running on some tegra SOC years ago... Maybe someone can start a froronix, which tests hardware with 100% free software and instead of bechmarks (which would be nice too) publishes checklists of whether it boots, display works, network works, etc. Pity it would likely require much more work building the platform support. Well, last time I checked phoronix didn't accept SEPA donations and I payed nothing, so I may not be the best one to propose anyone to spend effort on things...

[mr4kino]

Hi guys,

Noob question here: As this is running Linux and uses Arm we should be able to run a Kubernetes cluster?

Many thanks

[AndyChow]

I'm buying this. I hope that CUDA-X fixes the current ecosystem mess. Even installing TensorRT is painful right now.

This is a platform to test and run AI stuff. If you don't, there are plenty of more attractive solutions.

I'd be curious how this compares to Google Coral dev board. And that platform is even more blob-like than nVidia. You have to use an online service to convert your models to tensorflow-lite!

That said, this year is really interesting for AI. Can't wait for Google's Coral PCIe card which arrives in a few months.

[andreano]

Not unlike the Odroid N2.