Which tv box with Amg soc currently supports 4: 2: 2 H264 or H265 videos either by hardware or software? S922x?
i tried a rpi4 and it doesn't work with 4: 2: 2 H264 or H265 videos
Very few boxes support 4:2:2 video as it isn't a consumer format (DVD, DVB TV broadcasts, Blu-ray, UHD Blu-ray, Netflix, Prime etc. are all 4:2:0) so there is no reason to implement it in chipsets aimed at consumer devices.
4:2:2 h.264/h.265 is only used by broadcasters on contribution circuits - not final-leg distribution - so unless you are building a box for feed hunters (a small market) 4:2:2 isn't really a 'must have' feature.
None of the AMLogic chipsets that I am aware of support 4:2:2 hardware decode for this reason - so you need a box fast enough to decode it (and if it's interlaced also YADIF or W3DIF 2x deinterlace it) with CPU power.
The only ARM media player platform that I know that can currently do software decode (and deinterlace) of 1080i25 for MPEG2 and h.264 4:2:2 is the Apple TV 4K (the ARM SoC in that is a beast)
The other BIG advantage of the Apple TV 4K is that it has hardware acceleration support for 4:4:4 and 4:2:2 h.265/HEVC decode - because Apple use that codec for their Airplay/Sidecar iPad dual display function which compresses desktop video to h.265 for carriage over WiFi or Lightning/USB Type-C connection (to avoid reducing the chroma res to 4:2:0 and it all going smeary when you use your iPad as a second display) The Apple TV also has that functionality. I've played 2160p50 4:2:2 h.265/HEVC on the Apple TV in mrMC with very low CPU (but it was still not perfect because of some A/V sync issues)
If you want MPEG2 and h.264 4:2:2 with LibreElec then a decent Intel or AMD solution is probably your best bet - I used to use a Haswell i5 NUC to play 4:2:2 stuff and it coped - just - once multithreaded software decode of h.264 was implemented in ffmpeg. However the LibreElec implementation seemed to do a 4:2:2 to 4:2:0 conversion without interlaced aware chroma decode/conversion so you had saturated areas deinterlaced with p25 rather than p50 motion (a known issue in ffmpeg)