So yes and no on that recommendation. If you are just hosting content for local consumption, transcoding is unnecessary since you have the network bandwidth to just throw the data directly to whatever is playing it. So weaker hardware is perfectly fine. If you are doing lots of concurrent streams or there is network access outside the house, the limited bandwidth can become an issue so transcoding suddenly matters and more powerful hardware comes into play.

I have used many ARM SBCs and a few low-power Intel boards like my current N100 and they’ve all been fine. While I generally dislike Intel their quicksync is very useful in media server configurations. If you are going to be doing a lot of live transcodes, I would consider throwing an ARC GPU in there and having jellyfin utilize the transcode capabilities of the Intel GPU instead of the CPU as it can handle more simultaneous streams. Beware the xe driver as there are issues with it in certain configurations. Same with HuC/GuC. The older standard driver is more likely to just work. Jellyfin and the archlinux wiki have great documentation on this.

NVIDIA used to be top tier here but their transcode tech is pretty old by this point and the quality, while acceptable, isn’t the best. Intel beats them. AMD, generally a preference for me, has a terrible media transcoder. Easily the worst quality of all of them. For raw compute and pushing pixels, AMD all the way but for transcode I would pass.

So to summarize: cheap out if it’s just local access. Transcode is pretty much unneeded. If it’s outside the home and/or had many streams at the same time, Intel for the GPU and AMD for the CPU.