VidCoder h.264 QSV vs. NVENC

[Allan74]

For the last while I have been trying to find a more efficient way to shrink large 1080p MKV files for media server storage and keep them @ 8196kbps/AC3, before I make the switch entirely to h.265.

As I was told on this very forum, if a powerful enough CPU wasn't an option, QSV would give me the best image quality results, so that is out of the way.

I did however do a direct comparison using NVENC, which in VidCoder is limited to single pass work, benchmarking it against QSV in a single pass with all of the same basic settings set to the highest quality file possible.

I found that NVENC is obviously faster on it's highest quality single pass compared to QSV (on my hardware), but the NVENC 8192kbps file is also about 2GB Smaller (8GB vs. 6GB movie file).

Is this smaller NVENC encoded file a simple result of lesser quality, even at the same resulting bitrate ? or am I missing something else ?

Using the same input file, I also was surprised to have found that the QSV h.264 encoded file is the exact same size after either a 1 or 2 pass run.

NOW, I am going to stick with QSV 2pass, but am wondering if I am doing something wrong, seeing as 1 or 2 pass result in the same file size regardless and how the NVENC file is much smaller in size at the same bitrate when comparing even single pass runs.... ?

thanks in advance.
Allan

ps. Geforce GTX 950 vs. i7 7700k.

[sneaker]

file size = bitrate * duration
If the file size is smaller then the bitrate is also smaller.

[Allan74]

Not in THIS case.....
Same 12GB h.264 input file. Same Encoding Software. Same Codec. Same set bitrate (8192). Both Single Pass runs, set to highest quality possible for each test.

NVENC Single Pass resulted in a 6GB File @ 8192kbps
QSV Single Pass resulted in an 8GB File @ 8192kbps.

[Allan74]

I just ran the test on a file that I had previously transcoded and shrunk it further and although it didn't make a complete liar of me, the results were much closer, where some could argue the difference in final size could be in the realm of margin of error, but even this time, it's clear that with the settings available to me within VidCoder, on the highest quality settings for each, the NVENC h.264 Encoder shows a clear lead in compression, unless this is the result of dropping info and resulting in a lesser quality file...?

I had planned to post screenshots of all of this, but it would only lead to controversy and me having to further defend the finding.

I also believe that if I start out with a 'virgin' source instead of something that I have transcoded already, the results would be as I had stated above, with a greater discrepancy between the end resulting file sizes.

Does the NVENC offer better compression over QSV or worse source retention during the transcode ?

[jagabo]

file size = bitrate * duration
If the file size is smaller then the bitrate is also smaller.

Not in THIS case.....

In all cases (aside from minor container differences, muxing options, audio size, etc.). One encoder (or maybe both) didn't deliver the requested bitrate.

[sophisticles]

Not in THIS case.....
Same 12GB h.264 input file. Same Encoding Software. Same Codec. Same set bitrate (8192). Both Single Pass runs, set to highest quality possible for each test.

NVENC Single Pass resulted in a 6GB File @ 8192kbps
QSV Single Pass resulted in an 8GB File @ 8192kbps.

Not possible, the only way this happened is of the NVENC was an incomplete encode, i.e. there are frames missing.

Alternatively, the software you are using to determine bit rate of final encode is misreporting the bit rate.