CUDA encoding

[BlurayHD]

Hi Guys I just cant understand one thing, in some post some one told me encoding video with Nvidia CUDA final quality is worst than you encode with software, so, thinking about it AI use CUDA isnt? or you need a graphic card, so the video quality is not so bad?

Thank You in advance

[Subtitles]

Hi Guys I just cant understand one thing, in some post some one told me encoding video with Nvidia CUDA final quality is worst than you encode with software, so, thinking about it AI use CUDA isnt? or you need a graphic card, so the video quality is not so bad?

Thank You in advance

I use CUDA with Nvidia graphic card and it helps me with Whisper AI. I am not sure if it works without a GPU.
I Also use the graphic card to convert (encode) videos to different extensions and the result is excellent.

[pandy]

Normally video on NVidia cards is encoded by NVenc SoC - this is dedicated video encoder block not related directly with CUDA blocks - i recall in past there was some effort to create CUDA based video encoder but probably all work on this was abandoned years ago...

[frank_footer]

Hi Guys I just cant understand one thing, in some post some one told me encoding video with Nvidia CUDA final quality is worst than you encode with software, so, thinking about it AI use CUDA isnt? or you need a graphic card, so the video quality is not so bad?

Thank You in advance

the answer to your question is a bit convoluted.

cuda is framework developed by nvidia for creating programs written in fortran, c++, python and pascal to run either partially or completely on the gpu cores of nvidia graphics cards.

when the first cuda sdk was released by nvidia years ago they included lots of sample code to show how you could do various things with cuda.

one of the things they included was a bare bones h264 encoder that leveraged cuda to run on nvidia gpu cores.

this came to be known as the cuda encoder.

it was not meant to be a complete encoding solution, only a template for developers to build on, but that didn't stop people that wanted to cash in from compiling the bare bones encoder, slapping a gui on it and selling software.

Some of the software included badaboom, imtoo and xilisoft.

and those implementations were ok, but worse than the software encoders,

main concept has a better cuda encoder but still not great and there was a company that made a gpu encoder that was supposedly quite good, elemental i think they were called.

after intel came out with qsv, people basically abandoned pure gpu encoders, and nvidia and amd followed intel's lead and developed encoders based around dedicated encoding chips.

the thing is none of these is a pure hardware encoder unless you use them in so-called fixed function, low power mode.

under normal operation the encoder is partially hardware chip, partially cpu and partially gpu cores.

in the case of nvidia, if you are encoding in rgb color space, then everything is done on the gpu cores.

to tie this into your question, cuda ai can be as good or as bad as any other program depending on the skill of the programmer.

as for the current state of video encoding use video cards, there are many people that only use that exclusively.