NVidia Jetson Nano

[ible]

looks interesting -- quadcore cortex A57 and 128 Maxwell cores -- 5 to 10 W budget....
https://liliputing.com/2019/03/nvidia-jetson-nano-is-a-tiny-ai-computer-for-99-and-up.html
https://developer.nvidia.com/embedded/buy/jetson-nano-devkit

i forget what sort of consumption the OMAP5 draws, and whether there are some design requirements used by NVidia (i.e. they have to do the design in-house) for a future successor to the pyra.

 

[levi]

I don't know off the top of my head, but it's for sure a long way south of 10W. I remember a tale from when they were inventing the first ARM chip at Acorn; they plugged it in and started it running, but their ammeter didn't register any current. This being 1986-ish, I'd expect that to be an analogue multimeter kind of thing, but you could still see 10s of mW on those things pretty easily. That they couldn't see anything and though that was a single core chip clocked at a mere 8MHz with no kind of caching going on, it was still pretty remarkable.

For a more modern benchmark, I think the Pandora as a whole drew up to about 1W while running, including lighting the screen and all the other gubbins. Someone with a working unit can see this relatively easily using _wb_'s excellent Pandora System Info pnd.

 

[Chip]

Just checked my first-batch pandora with the system info pnd. It idles at just about 1W, and hits 1.8W under load. That's for the whole system with the display at max brightness and wifi on.

5-10W is just too high for a handheld device. Hell, the atom in the GPD win is only a 2-4W part, and that requires active cooling.