I’m running CoreELEC 188.8.131.52 on both my Minix U9-H (S912) and my new ODroid N2+ (S922X). I would describe the U9-H’s deinterlacing as “dodgy”. When it locks onto the cadence properly it’s decent, but it often briefly mistakes 50i content for 25p content, resulting in some artefacts when the content shifts from mostly stationary to mostly motion. However, having used the N2+ for a few days, I’d describe its deinterlacing as “bad”.
For example, when watching an tennis on an SD (576i) channel, the U9-H struggles to show the far base line when the camera moves, but the ball in motion is smooth. With the N2+, the far base line is rendered better but there are combing artefacts on the ball and moving raquets nearly all the time. When using the deinterlacer on my AMD Radeon Vega 56 (using MPC-HC + MadVR), none of these problems exist. I know that I shouldn’t expect the Amlogic chips to deinterlace as well as my GPU but it doesn’t make sense to me that the S922X looks worse than the S912 in this regard.
Surely the S922X wouldn’t have a worse deinterlacing algorithm/chip than the S912, so I don’t understand why it looks worse. Aside from the chipset, the only difference I can think of between the two boxes is the N2+ uses a newer kernel (amlogic-ng build) - maybe this is causing a problem? Are there any settings that control this?