Why don't I see a difference between H.265 and H.264 with CQP?

[Panossa]

Hello, I recently built a new PC with an RTX 4080 Super and I thought of using H.265 since I have "spare" resources now.

Settings I've used previously for recording (to use in editing software):

• 1080p
• H.264 NVENC
• CBR with 30k bitrate
• Preset P7, Tuning "High Quality", and Profile "high"
• Look-ahead off, Psycho Visual Tuning on
• 60 FPS (CFR)

Settings I've tried today:

• 1080p
• H.264 NVENC
• CQP with quality set to 26
• Same preset, tuning, profile, look-ahead, psycho visual tuning, fps (CFR).

and:

• 1080p
• H.265 NVENC
• CQP with quality set to 26
• Same Preset, Tuning, look-ahead, psycho stuff.
• Multipass Mode: Two Passes (Full Resolution)
• Profile: main

The problem(s):

1. A recording of two minutes of roughtly the same content (sadly no benchmark in Hunt Showdown, I just ran the same path) gave me video files with 40k bitrate on both of the new setups (CQP 26) with both encoders. Not even 1k bitrate difference. The files were also identical in size despite using H.264 vs H.265 and being of same length with the same average bitrate. Why?
2. I don't see a difference in quality. I would've assumed H.265 with the same bitrate and file size would have better quality. Why?

Also, is there a difference between rescaling in the "Video" tab of the settings (where the canvas settings are) and the Output -> Recording settings? I'm playing on a 1440p monitor but I want to record in 1080p. Currently using the "Video" tab for that with Lanczos)

(My goal is to record mostly fast-paced 3D games. If CQP is not the way to go despite what I've read here a few times, please tell me.)

 

[prgmitchell]

I think the size difference wouldn't be much depending on the content and you shouldn't expect a quality difference like this, you are targeting the same CQ level so the resulting video would be the same quality.

 

[Panossa]

I think the size difference wouldn't be much depending on the content and you shouldn't expect a quality difference like this, you are targeting the same CQ level so the resulting video would be the same quality.

I would expect to see a difference in either file size or visual quality between H.264 and H.265 but there was none in both and that's kind of confusing since H.265 is hailed to be superior. Afaik, CQP only means the quality stays the same but H.265 should be able to get the same quality at a lower bitrate, which it doesn't.

 

[koala]

As far as I see, you're able to increase the CQ value for h.265 to reduce file size, while not suffering from the same quality decrease as it would be the case with h.264. So you cannot directly compare CQ values. It's a difference of about 3 (which means half the file size). So if you use CQ=23 for h.264, you can use CQ=26 for h.265 to achieve about the same quality at half the file size.

You determine equivalent CQ values, you need to create videos with the same perceivable quality, make sure it's the same quality with post-analysis such as psnr or vmaf, then see what different CQ values you used to create the videos. I'm sure people already did this, however I don't have proper Google search terms at the moment. This is probably also dependent on the actual footage.

 

[prgmitchell]

I would expect to see a difference in either file size or visual quality between H.264 and H.265 but there was none in both and that's kind of confusing since H.265 is hailed to be superior. Afaik, CQP only means the quality stays the same but H.265 should be able to get the same quality at a lower bitrate, which it doesn't.

yeah sorry, I think the only issue here is inaccurate expectations.

 

[Panossa]

As far as I see, you're able to increase the CQ value for h.265 to reduce file size, while not suffering from the same quality decrease as it would be the case with h.264. So you cannot directly compare CQ values. It's a difference of about 3 (which means half the file size). So if you use CQ=23 for h.264, you can use CQ=26 for h.265 to achieve about the same quality at half the file size.

You determine equivalent CQ values, you need to create videos with the same perceivable quality, make sure it's the same quality with post-analysis such as psnr or vmaf, then see what different CQ values you used to create the videos. I'm sure people already did this, however I don't have proper Google search terms at the moment. This is probably also dependent on the actual footage.

This is very helpful, thank you!
Although I'm still not sure whether CQP or CBR is better for editing. I heard it's recommended to use CBR for performance in editing software, but that's probably irrelevant with a good rig.