Page 1 of 1

Free vs Paid .MP4 H.264 performance

PostPosted: Thu Oct 11, 2018 4:02 pm
by eevnos
Hey all,
I'm sorry if this has been answered, but I've been trying to find this info and have not been able to find concrete information.

Here's the details, if you want to skip to the question it's down below:
I have a Lumix G7 camera and shoot 4k, H.264, .MP4 files @ 24fps.
Now I know this is a HORRIBLE combination for video editing but please hear me out.

I have a custom built PC with an i7-8700K processor, 16GB or RAM and a GTX1070 GPU, running Windows 10.
Premiere runs perfectly, with the files mentioned above.. scrubbing, playback, Fusion, all work great with no optimized media needed.

I don't want to use Premiere anymore, I HATE the subscription model and LOVE how I pretty much get one stop shopping in Resolve. ;)

Resolve (free version) on the other hand does not work as well on this machine.
Scrubbing is really bad and playback is choppy and not the full 24fps.
I have similarly spec'd laptop that has very similar performance.
Settings don't matter, I've tried every combination you could think of, over the past several months, and nothing makes a significant improvement.

Now here's where it get interesting (to me anyway).. I have a 2013 27" iMac with a 4th gen i7 and GTX 780m GPU and the free Resolve 15 works GREAT with the files mentioned above (no optimized media).
Scrubbing, playback, fusion, all work great!
I would use this as my primary editing machine, but the desktop and laptop are much more powerful (from a strictly spec perspective) so would like to use those to speed up processing time in fusion and render times.

OK, so here's the question(s):
Is the Studio version more efficient with H.264, MP4 files on a Windows based machine?
If so, why?
From what I've been able to piece together, in the free version the decoding is limited to CPU based only and will not take advantage of the hardware decoder on the GPU, but the studio version DOES take advantage of this.. is that accurate?

I have another thread about why the Mac version works better, on a much lower spec'd machine, so won't go into that too much, but what I'm guessing is that the decoding available on the mac takes advantage of the GPU decoding by default.

Thanks for reading this somewhat long post.
I LOVE resolve and have no problem paying $300 for it, but I want to make sure that there really will be a performance improvement with the paid version.

*** UPDATE ***
The Studio version does indeed work MUCH better with H.264 MP4.
Everything is working PERFECTLY with studio version on the same hardware!

Re: Free vs Paid .MP4 H.264 performance

PostPosted: Thu Oct 11, 2018 4:35 pm
by Cary Knoop
What I would advice is to transcode H.264 video sources to a more editor-friendly CODEC.

Re: Free vs Paid .MP4 H.264 performance

PostPosted: Thu Oct 11, 2018 5:40 pm
by eevnos
Cary Knoop wrote:What I would advice is to transcode H.264 video sources to a more editor-friendly CODEC.


While that is a workaround, it just adds too much time to the overall production time when there is other software available that doesn't require that no the same hardware.
Also, it doesn't answer my question. ;)

I'm wanting to know if the claims that H.264 performs better in Studio are accurate.

Re: Free vs Paid .MP4 H.264 performance

PostPosted: Thu Oct 11, 2018 5:44 pm
by Gary Hango
On Windows I believe the free version uses the OS supplied h264 decoder which doesn’t use GPU decoding. The Studio version installs its own h264 decoder with GPU accelerated decoding, so the Studio version should fly with you clips. Can others please verify?

Re: Free vs Paid .MP4 H.264 performance

PostPosted: Thu Oct 11, 2018 5:57 pm
by Cary Knoop
eevnos wrote:
Cary Knoop wrote:What I would advice is to transcode H.264 video sources to a more editor-friendly CODEC.


While that is a workaround,....

It is not a workaround, it is a sound workflow!

Re: Free vs Paid .MP4 H.264 performance

PostPosted: Thu Oct 11, 2018 6:00 pm
by Carsten Sellberg
eevnos wrote: Is the Studio version more efficient with H.264, MP4 files on a Windows based machine?
If so, why?


Hi.

As I understand do both the Windows and the Mac free version use the OS buildt in H264 to decode. May be the Mac version is better?

But on the paid Windows Studio version is one format of H264 decoded on the GPU by this API:

https://developer.nvidia.com/nvidia-vid ... ECFeatures

As I understand it is it only 8 bit and only YUV 4:2:0

Regards Carsten.

Re: Free vs Paid .MP4 H.264 performance

PostPosted: Thu Oct 11, 2018 9:44 pm
by eevnos
Thanks Gary and Carsten, this is what I've been able to piece together as well.

I think the MacOS one works so well because the default H.264 decoder utilizes the GPU hardware decoder.

It's just interesting that I have not been able to find any reference to that in the literature from Blackmagic.
I have found reference to improved ENcoding, in Studio, but not DEcoding.

I think I'm just going to go for Studio and keep my fingers crossed. ;)

I'll post back if there's an improvement.

Re: Free vs Paid .MP4 H.264 performance

PostPosted: Thu Oct 11, 2018 9:53 pm
by Dan Sherman
Gary Hango wrote:On Windows I believe the free version uses the OS supplied h264 decoder which doesn’t use GPU decoding. The Studio version installs its own h264 decoder with GPU accelerated decoding, so the Studio version should fly with you clips. Can others please verify?


This is partially correct, The studio version doesn't install anything, it use the video decoding api built into Nvidea graphic cards.

https://developer.nvidia.com/nvidia-video-codec-sdk

Re: Free vs Paid .MP4 H.264 performance

PostPosted: Thu Oct 11, 2018 9:55 pm
by Dan Sherman
eevnos wrote:I think I'm just going to go for Studio and keep my fingers crossed. ;)


I have a G7 and the studio version running on a 6850K/1070 combo. You will be perfectly fine.

Re: Free vs Paid .MP4 H.264 performance

PostPosted: Sat Oct 13, 2018 11:46 pm
by eevnos
Just wanted to report back that I got my Studio license key today and Studio works PERFECTLY on my hardware!

Anyone wondering if Studio will work better with H.264 MP4 files, the answer is YES.

It appears that the default H.264 decoder (that is supplied with Windows) does not take advantage of hardware decoding where as the decoder supplied with Studio does.

Everything is very fluid and plays back at full speed on my same hardware, DEFINITELY worth $300!

Re: Free vs Paid .MP4 H.264 performance

PostPosted: Sun Oct 14, 2018 12:22 am
by Marc Wielage
eevnos wrote:Just wanted to report back that I got my Studio license key today and Studio works PERFECTLY on my hardware! Anyone wondering if Studio will work better with H.264 MP4 files, the answer is YES.

That's good news!

I still think it's a better workflow in general to transcode, because H.264 does tax the CPU a bit more and ties up more resources. My general opinion is that Long-GOP formats (like H.264) are not great for post; they're intended more as delivery formats (like YouTube and Vimeo).

Re: Free vs Paid .MP4 H.264 performance

PostPosted: Sun Oct 14, 2018 12:44 am
by Chad Capeland
Marc Wielage wrote:
eevnos wrote:Just wanted to report back that I got my Studio license key today and Studio works PERFECTLY on my hardware! Anyone wondering if Studio will work better with H.264 MP4 files, the answer is YES.

That's good news!

I still think it's a better workflow in general to transcode, because H.264 does tax the CPU a bit more and ties up more resources. My general opinion is that Long-GOP formats (like H.264) are not great for post; they're intended more as delivery formats (like YouTube and Vimeo).


Might just be reviewing footage to make selects to transcode. Transcoding all of your footage might take longer than just the selects, depending on the shoot. I don't even know if Resolve is the right tool for doing the review and transcode, as I shoot DNG, but I could see someone trying this.

Re: Free vs Paid .MP4 H.264 performance

PostPosted: Sun Oct 14, 2018 1:20 am
by Dan Sherman
Marc Wielage wrote:I still think it's a better workflow in general to transcode, because H.264 does tax the CPU a bit more and ties up more resources. My general opinion is that Long-GOP formats (like H.264) are not great for post; they're intended more as delivery formats (like YouTube and Vimeo).


This is not correct in all scenarios.

If you have Resolve studio version 14 or newer, and a Nvidia Kepler or newer gpu, then 8bit h.264 will be decoded by the gpu. Thus putting zero load on the CPU.

Re: Free vs Paid .MP4 H.264 performance

PostPosted: Sun Oct 14, 2018 8:19 am
by Marc Wielage
Chad Capeland wrote:Might just be reviewing footage to make selects to transcode. Transcoding all of your footage might take longer than just the selects, depending on the shoot. I don't even know if Resolve is the right tool for doing the review and transcode, as I shoot DNG, but I could see someone trying this.

CinemaDNG is a fantastic format and I never transcode that. It works fine on my system and is not based on H.264 -- it's a frame-based format.

Dan Sherman wrote:This is not correct in all scenarios. If you have Resolve studio version 14 or newer, and a Nvidia Kepler or newer gpu, then 8bit h.264 will be decoded by the gpu. Thus putting zero load on the CPU.

It's still a big load on the system and will bog things down unnecessarily. Newer formats like H.265 and HEVC are quite a bit cleaner, and for those I don't suggest you transcode if you can beef up the system to handle it. But every time a client brings me stock footage, DJI drone shots, GoPros, and stuff like that specifically using Long-GOP H.264, we always insist on it being transcoded so as not to create workflow issues. If it works for you, fantastic -- but we won't do it at my facility.

Transcoding is fast and easy and can be handled in batch mode by many utilities; even a second copy of Resolve on a separate system can crank out identically-named files with matching timecode that will easily relink into a conformed master. I sometimes will use some mild NR as part of the transcode process, which helps cover up any obvious H.264 macroblocking issues.