DaVinci Resolve H.265: M2 Max Exports Smaller, Lower Quality

Hey everyone, I've got this really strange issue in DaVinci Resolve and I'm hoping someone here might be able to point me in the right direction.

Problem:

My friend and I are both using Mac Studios to edit. I have an M1 Max, and he's got the newer M2 Max. The thing is, the same exact project with identical export settings produces vastly different results. On my computer, I get an expected size of 2.04GB. while with identical settings he gets 474MB.

Troubleshooting So Far:

We've triple-checked all our export settings (Even made a preset to ensure identical settings)
Source footage is the same on both machines.
Using the same DaVinci Resolve version

Screenshots:

Screenshot 2024-03-21 at 9.00.25 am.png (129.46 KiB) Viewed 1221 times

Screenshot 2024-03-21 at 9.00.19 am.png (140.8 KiB) Viewed 1221 times

Screenshot 2024-03-21 at 9.00.14 am.png (131.29 KiB) Viewed 1221 times

Anyone seen something like this before? Any ideas?

Here's some more info that might help:

DaVinci Resolve Version: 18.6.4
Source Footage Details: RED Raw footage 6144x3240
Specific Export Settings: H.265 4:22 10bit, Source Resolution

Any help would be much appreciated.

Even if your settings are the same, there is a difference:
You both checked
"Use hardware acceleration if available"
and as you mentioned, your hardware differs.
What are the results, when you disable hardware acceleration ?

Software encoding should always deliver the same results.
Hardware encoding depends on the graphics card used and,
as mentioned in many articles, can lead to poorer quality.

Even if your settings are the same, there is a difference:
You both checked
"Use hardware acceleration if available"
and as you mentioned, your hardware differs.
What are the results, when you disable hardware acceleration ?

Thanks for the suggestion,
Unfortunately, I cannot render h.265 with hardware acceleration disabled. I've just Run the test using H.264 without hardware acceleration and am getting the identical file sizes for both computers (4.62GB). Is this an Apple Silicon issue, or a davinci issue?

I'd much prefer to use h265 over h264, as it allows me to render 10bit, as well as higher quality for a smaller file size. I'm using these renders as an intermediate from the RAW, to edit, and file size efficiency is super important due to server space considerations.
Thanks

I've done some further testing that might be helpful:

I rendered the RAW source footage as ProRes4444 in DaVinci Resolve. I then used Handbrake to export both versions (M1 and M2) of the ProRes4444 file using handbrake's hardware-accelerated HEVC encoding. They both came out identical.

Is there any settings in Resolve that can fix this?
Has anyone else experienced similar issues with DaVinci Resolve exports on M2 Max Mac Studios?
Thank you!

I'm not an expert but noticed two things:

Data Levels: Auto (the PC is making decisions based on ????)

Quality: Optimize for speed (at what expense?)

I'm using these renders as an intermediate from the RAW, to edit

I think you're better off accommodating your storage needs and editing the RAW footage.

I'm not an expert but noticed two things:

Data Levels: Auto (the PC is making decisions based on ????)

Quality: Optimize for speed (at what expense?)

Thanks for the suggestion. I've just set Data Levels to Full and also removed Optimise for speed. I still get a drastic difference in size (4.57Gb on the M1 and 609MB on the M2)

I think you're better off accommodating your storage needs and editing the RAW footage.

As much as i'd like to, we need to have footage accessible in the future and with the amount of data we have, keeping RAW's for every interview is unfeasible. (By the way I'm only testing on a 5 minute segment from one of the cameras, a full RAW camera is upwards of 500GB, which i hoped to render to around 50GB. on a 3cam Multicam setup the storage costs get out of control, Not to mention the need for easy playback and editing for collaborators with slower machines)

I do appreciate any help on this matter, so thanks.

[Dwaine Maggart]

If you examine the files using MediaInfo, are there any obvious differences there?

What do you mean you can't export H.265 with hardware acceleration disabled? What happens?

Isn't using h.265 or h.264 for editing frowned upon as they are more delivery codecs?

Well, at least H.265 can be Intra only, 10 bit, 422. If you up the bitrate and have the right hardware, it can serve as an intermediate. But, if I may ask, why stress the machine with it?

If you examine the files using MediaInfo, are there any obvious differences there?

What do you mean you can't export H.265 with hardware acceleration disabled? What happens?

No difference in metadata as far as I can tell. I get this Error message when doing a h.265 Main 10 422 export,

Screenshot 2024-03-22 at 08.22.19.png (18.6 KiB) Viewed 649 times

But I've just tried it again with Main 10 and it seems working fine. The renders are going slow so I'll post those results when they are done.

Isn't using h.265 or h.264 for editing frowned upon as they are more delivery codecs?

True, it's not an ideal situation, but these are for YouTube and previously the system was far worse, relying on very low bitrate h.264. I've been trying to find a good intermediate codec that will still look great, even if it doesn't perform as well as RAW in grading. I've settled on H.265 as at 120,000kb/s it offers about 1/10th the size of the RAW file and looks good.

Well, at least H.265 can be Intra only, 10 bit, 422. If you up the bitrate and have the right hardware, it can serve as an intermediate. But, if I may ask, why stress the machine with it?

The main reason is storage. At the company I work for, we do 2 full 1hr30 to 2hr podcasts every week. We need to keep the files easily accessible to a small team (so they can cut clips for social media etc...) If we were to keep the RAW, it would A, be very expensive storage wise, and B not be ideal for everyone on the team.

I am happy for any suggestions for alternative codecs. In my testing I've found ProRes 422 at 4k to be only half the size of RAW.

Here is my testing:

Screenshot 2024-03-22 at 08.43.24.png (49.19 KiB) Viewed 649 times

Thanks!

With Hardware encoding off H265 Main 10, the renders come out identical. But it took over 2 hours to render a 5 minute clip, so it's not a solution that works for me. Given all the testing, I am lead to believe that this issue is with Davinci Resolve's Hardware acceleration on specifically the M2 chip specifically for h.265. Or a computer specific issue

Here's why:
- No issues on the M1 (files are as expected)
- A hardware accelerated handbrake encode results in the same file as the M1
- A software encode on Davinci results in the same size file (4.57Gb on both)
- A hardware accelerated h.264 encode results in the correct expected size.

I am open to alternatives and suggestions for what I can do to work around this issue. My only ideas are to:
- Render everything on the M1
- Render a Prores Intermediate, then re encode with handbrake
- Find another Codec that works. (I am tempted to use h264, but I'm not happy about it... )


Can anyone else confirm this behavior? Thanks.

What is actually the problem? Encoding H.265 using various kinds of hardware is less efficient than using software encoding. But as long as the bitrate is high enough you are not compromising quality you just get a larger file.

What is actually the problem? Encoding H.265 using various kinds of hardware is less efficient than using software encoding. But as long as the bitrate is high enough you are not compromising quality you just get a larger file.

You're right, hardware encoding can be less optimized for file size than software encoding. However, the issue I'm facing is that the M2 Max, using hardware encoding for H.265, produces a much smaller file size compared to the M1 Max or even software encoding on the M2 Max. This smaller file size is representative of the quality of the file (the output from the M2 is visually lower quality). In my current workflow I would like both the M1 computer and the M2 computer to produce the same files.

Software encoding takes considerably longer, and I have no problems with the outputs of a hardware encoding on the M1, I would just like both computers to have the same output.

What is actually the problem? Encoding H.265 using various kinds of hardware is less efficient than using software encoding. But as long as the bitrate is high enough you are not compromising quality you just get a larger file.

You're right, hardware encoding can be less optimized for file size than software encoding. However, the issue I'm facing is that the M2 Max, using hardware encoding for H.265, produces a much smaller file size compared to the M1 Max or even software encoding on the M2 Max. This smaller file size is representative of the quality of the file (the output from the M2 is visually lower quality). In my current workflow I would like both the M1 computer and the M2 computer to produce the same files.

Software encoding takes considerably longer, and I have no problems with the outputs of a hardware encoding on the M1, I would just like both computers to have the same output.

Have you tried raising the bitrate? Also play with disabling the "Optimize for speed" option.

Have you tried raising the bitrate? Also play with disabling the "Optimize for speed" option.

I hadn't tried raising the bitrate seperateley as I was under the impression that both computers would produce similar outputs with the same export settings. But i've just tried it at 900,000 kb/s and the output is identical to that of 120,00kb/s. Strange.

and yes, optimise for speed is off.

Have you tried raising the bitrate? Also play with disabling the "Optimize for speed" option.

I hadn't tried raising the bitrate seperateley as I was under the impression that both computers would produce similar outputs with the same export settings. But i've just tried it at 900,000 kb/s and the output is identical to that of 120,00kb/s. Strange.

and yes, optimise for speed is off.

That means that you already have a bitrate of a given quality and anything more will not make it any better.. One of the bummers in Resolve is that there is no CRF option (but in the interest of fairness some hardware encoders do not have that). There are other quality options, for instance QP. But I would want to see an engineer argue that QP is of any practical use for video production.

That means that you already have a bitrate of a given quality and anything more will not make it any better.. One of the bummers in Resolve is that there is no CRF option (but in the interest of fairness some hardware encoders do not have that). There are other quality options, for instance QP. But I would want to see an engineer argue that QP is of any practical use for video production.

Interesting, is is this a hardware issue then? When doing a handbrake Hardware encode this issue isn't present.

Interesting, is is this a hardware issue then? When doing a handbrake Hardware encode this issue isn't present.

Handbrake can use different encoders. x265 is arguably one of the best but it is also a software encoder. Using a hardware encoder through Handbrake will give the same limitations. NVENC, which is NVIDIA's encoding technology does not have a crf option (for crf you need a lot of inter-thread communication and synchronization which is very dynamic and hard to implement for a hardware model).

Would Anyone with an M2 Max mac studio be able to verify this isn't a one off issue that i'm having please? exporting anything with a bitrate of 100,000 on h265 422 and the same in h264 and compare the files? Thanks.

My friend and I are both using Mac Studios to edit. I have an M1 Max, and he's got the newer M2 Max. The thing is, the same exact project with identical export settings produces vastly different results.

What specific format did you shoot on for production?

Call me crazy, but we tell our clients not to use H.264 (or H.265) for offline editing. If you're both on Macs, just export to ProRes 422LT (or if you want even smaller files, ProRes Proxy). If you're dealing with 4K files, go with exactly half resolution, same frame rate, same aspect ratio. This is guaranteed to work on Macs and will still have embedded timecode, and you can later relink to the original files with no problem.

I've seen ProRes 422LTs actually hold up pretty well for preview screenings and network previews, so it's not an awful format.