[noggin]

If you want to be fully complete - then you can also consider that the HDMI 2.0 4:2:0 8-bit YCbCr 2160/50-60p mode can be carried at the lower, HDMI 1.4b clock rate (which is how NVidia and Sony were able to upgrade graphics cards and TVs with HDMI 1.4b physical hardware to support this bit of HDMI 2.0)

Agreed, but that's why I stated "uncompressed" 4K@60p. 4:2:0 has compressed chroma, thus requires lower bandwidth, thus fits within HDMI 1.4 bandwidth.

Yes - though I think 'compressed' and 'uncompressed' are misleading terms. The terms used everywhere I encounter 4:2:2 and 4:2:2 are 'sub-sampled'. 'Compressed' is reserved for compression schemes like MPEG2, H264, AVCi100, DV100 etc. 4:2:2 has horizonally sub-sampled chroma, 4:2:0 has horizonally and vertically sub-sampled chroma. 4:4:4 YCbCr and RGB are equal bandwith with no sub-sampling. HDMI and HD-SDI are both considered uncompressed interconnects - as the video remains in the baseband domain and runs in real-time (i.e. you don't have to run it through frame stores to generate video)

If you describe baseband HD-SDI 4:2:2 as 'compressed' you'll get a lot of strange looks as it is usually, and specifically, described as 'uncompressed' in broadcast circles...

You only need the HDMI 2.0-specific higher bandwidth for the 2160/50-60p 4:2:2 and >8 bit 4:2:0 modes I believe. (There isn't a 4:4:4 RGB or YCbCr implementation for >30p at 2160p even with HDMI 2.0)

"4:4:4 RGB" is another misconception, because 4:4:4 (or 4:2:2 or 4:2:0, etc.) denotes luminance and chroma subsampling (=compression) of YCbCr color space. There's no such thing in RGB color space, RGB is always uncompressed (=fully sampled as opposed to subsampled).

It does now... But the 4:4:4 and 4:2:2 nomenclature has been with us since the 80s, when 3:3:3 RGB and 8:8:8 RGB were also used in some devices. The numbering scheme is all about relative sampling rates - and there have been RGB-based systems that use rates other than 13.5MHz (in SD).

For many using the same 4:4:4 for RGB and YCbCr sampling ensures it is clear that the same sampling rates are being used.

HDMI 2.0 can carry 4K@60p in RGB and YCbCr 444 at 8bpc. >8bpc@60p apparently requires compression.

Yes - I had missed out the 4:4:4 RGB/YCbCr 8-bit modes (I'd convinced myself they didn't exist!)

The >8 bit modes require subsampling to 4:2:2 or 4:2:0 and are YCbCr only.

(Bear in mind it's not a problem for 4K 10bpc movies as movies are running @24p and HDMI 2.0 can carry 4K up to 30p in RGB and YCbCR 444 at up to 16bpc)

No - but for broadcast purposes where you have to run at a fixed 50 or 59.94Hz rates that's a moot point.