Radxa Dragon Q6A--Qualcomm QCS6490 Octa-core Platform for Edge AI

Great purchase experience. Product, not so much

Not Araces fault at all, but the board has a lot of growing to do still. AI devel libraries and process are terrible. The latest OS build from Radxa itself does not correctly support required libraries for the AI stack, requiring a pile of workarounds, and even then the samples are OK, but to deploy your own is a huge PITA that you have to workaround, and you better have a fast desktop for the processing or dont even bother. I think the board will eventually be very good because of Qualcomm, but currently it's not worth the hassle for AI. if you just want a competent SBC, then it's fine.

Fast shipping and great price. The board still lacks some models for the NPU on qualcomm side. Very efficient and fast. Armbian is not fully working if you want to use the NPU atm, only radxa os

Very powerful and well-built

It is a great board for the price! It has great potential, very fast and quite standard due to being Qualcomm. The documentation could be better though. The HEVC HW encoder works and is accessible but there is not ready-made app like rpicam-vid. Libcamera is supported and can be used as a gstreamer source. But I can't figure out how to get more than 7 FPS of 4K video, the performance is probably lost in video conversions and I should be doing that in HW too but it seems to be difficult given not enough documentation. It has a huge potential but requires skills to extract it. It will work perfectly for other more normal tasks, including ML/AI. Overall I can recommend it.