10-bit DC Component Precision...? benefitial?

[plonk420]

in TMPG and HC, what exactly does this do? until i re-read the tooltip in TMPGEnc, i thought it referred to the input or output video having a 10-bit (per channel) color depth or however that's worded...

since that's not the case, i'm not sure i understand it. is it using any more bits (that could be better used keeping blocks down) than to set it to 8-bit?

[Boulder]

A higher precision will use more bits, but the difference is quite small IIRC. You can test it by encoding a sample clip in CQ mode and using different DC precisions while keeping the CQ parameter the same.

[foxyshadis]

It's an option to pull just a tiny bit more quality out when you've already maxed out the bitrate, because it reduces the roundoff errors inherent in the DCT transform. It's only useful when decoding roundoff causes errors <= 1-2 values. If you use it at a low bitrate all you're doing is storing artifacts at a higher resolution, which is pretty much useless, and robbing it of 1% or so bitrate.

CCE has an automatic setting that will increase it if the quantizer is low enough; their guideline is "8-bit below 4mbps, 9 at 4-8, 10 above 8", but like any bitrate recommendation it varies based on compressibility.

[MrTroy]

If I set 10 bit DC precision in TMPGEnc the colours become quite oversaturated. On other forums I've read that being described as "the colours become much more beautiful", but I don't think I like the saturation. ;)