Question / Help NVENC CQP 0 and 1 and lossless.

[RaceDriverMIKU]

Hello.
When I try to use CQP 0, Hi444P, High Quality, 2-Pass,I444 with full, bitrate is about 55~100 with high motion. (I used b frames with my last footage. but without b frames quality is same.)
CQO 1, Hi444P, Default, 2-Pass,I444 with full, bitrate is go beyond 200Mbps. (avarage? 266Mbps, no b frames)
lossless ,Hi444P, High Performance, 1-Pass,I444 wirh full, bitrate also go beyond 200mbps (avarage? 299Mbps), but at teast 1/3 frames failed to encode in time. (36~47% of frames, High Quality causes worse.)

My Question is, why cqp 0 has less bitrate than cqp 1? and why has worse quality than cqp 1?(It looks worse to me.)
Why performance of lossless rate control is really bad? but cqp 1 is just fine, only 2~3% of frames are failed.(2~3% is just only when start.)

I found nothing usful in log, but I put a link.
https://obsproject.com/logs/DyUYS1FaXHNz9Lkk

Thank you.

 

[koala]

Your GPU isn't able to encode all frames. It skips due to encoding lag. With cqp 0 more than with cqp 1. Since these skipped frames are not written to the video file, it's no surprise that a video with cqp 0 and 39% skipped frames is smaller than a video with cqp 1 and 3%.
I don't understand how you say 2-3% is fine - this is 1-2 frames every second. This is visible as laggy, sometimes choppy video. As best case you perceive it as non-smooth and it thwarts the purpose of 60 fps recordings, which should appear silk-smooth in comparison to 30 fps videos.

 

[RaceDriverMIKU]

Your GPU isn't able to encode all frames. It skips due to encoding lag. With cqp 0 more than with cqp 1. Since these skipped frames are not written to the video file, it's no surprise that a video with cqp 0 and 39% skipped frames is smaller than a video with cqp 1 and 3%.
I don't understand how you say 2-3% is fine - this is 1-2 frames every second. This is visible as laggy, sometimes choppy video. As best case you perceive it as non-smooth and it thwarts the purpose of 60 fps recordings, which should appear silk-smooth in comparison to 30 fps videos.

Hello. sorry for lack of information.
With cqp 0, not only less bitrate, but it is... smoother than lossless rate control.
Encoding failure happens at the beginning of recording, and when games load big data, not every seconds.
My short tests were 30~60 secs, and all of 2~3% failures are only at beginning.
Sorry for confusion.

bitrate
lossless rate control>cqp 1>>>>>>>>cqo 0
smoothness
cqp 0=cqp1>>>>>>>>>>lossless rate control
As theory I know, lossless rate control = CQP 0. but it isn't anyhow. it looks just performs different.
I really don't understand. So I guess there is something that users preventing using cqp 0 in code. and replace cqp parameter as 21.

 

[koala]

To be honest, that's something that's not important to me. It's not clear how lossless nvenc lossless actually is, so I don't bother with such low cqp values. If I need good quality, I use the usual values around 18 that are visually indistinguishable from the original but still produce reasonably good compression in comparison to lossless.

I once had a hard requirement for a lossless recording. It had to be lossless, every bit lossless. I didn't use nvenc for this but software instead, because the losslessness of lossless nvenc wasn't clear to me. OBS simple output mode uses ffmpeg for this with a a special codec. utvideo as far as I remember, and this is true lossness.