I am just wondering if anyone has tried this on other SBCs or other mini pc? I know other images were in the works but just curious if it’s feasible to compile on something using the Rockchip 3588s (Orange Pi 5 or Khadas Edge Pro, etc) or if I should get something like a little Celeron N5005 or i3 mini pc to tide me over for a little while. I do have a Jetson nano collecting dust and may try to compile everything on that but that is a serious project potentially and I do plan on getting a Jetson Orin early next year if they are actually available. Just curious what others have tried so far? What kind of update rates are you getting with your setups?
I was running the pr code on my nuc, fed from a rpi4 that had the kraken on it and all worked well
I’m in my infant SDR steps, but I chose ZimaBoard with Ubuntu, and so far, I’m a happy camper. It just works.
I use the 8GB/32GB version, feeding it both from the wall and the portable battery, never seen more than 7Wt over the 12V line, no crashes or lags.
Works OK on the Latte Panda 3 Delta.
For DOA almost any cheap device with power similar to a Pi4 works well.
We’ve been using Wyse 5070 thin clients as DOA test benches (sold cheaply on eBay as ex office equipment). The OPI4 works well too, but i’ve been having major issues with it not booting correctly all the time (completely unrelated to the Kraken code, something to do with the EMMC).
I have an OPI5 preordered for testing as well. That looks like a really great device hardware and price wise, at least until the Raspberry Pi 5 eventually comes out.
The Jetson Nano has also been tested and it works, but the CPU is a bit underpowered compared to a Pi4. The GPU is not used in DOA so there’s no benefit to dedicating such an expensive device to it.
If you’re doing stuff in GNU Radio, it’s better to use a more powerful device.
Could you use the GPU with some different numba decorators?
Yes or by using cupy, but I don’t think it’s as simple as just adding a decorator. Some work to write things in a GPU friendly way would still be required, and the time to invoke the GPU is probably not worth it for just DOA.