What algorithm is used in ME only mode in NVENC?


What algorithm is used in Motion Estimation only mode in NVENC (H.264 encoding)? Is it sad or sad + lambda*R?

Is it possible to program NVENC for using different algorithms? We develop H.264 encoder and we whould like to try different methods in ME, selecting mb_type, mb_part, sub_mb_part, refernce frames and etc. Is it possible to use NVENC for it?


1.The NVECNODE API doesn’t expose the ability to change the motion estimation algorithm.

2.By default NVENCODE API enables all partition types(16x16, 16x8, 8x16, 8x8), macroblock type(Inter/Intra). The support for partition types below 8x8 is not there. The reference frame selection should be done by client.

Interesting. So the algorithm used by nvenc is sad? not sad+lambda*R