1 Correct answer
If you have CUDA turned on, then no.
If you have CUDA turned off but aren't scaling any media, then still no.
21
Replies
https://forums.adobe.com/people/R+Neil+Haugen wrote
For most work, MRQ MBD aren't needed. Using them will slow down exports, and occasionally may even cause problems. Both have only a few uses these days. From a very useful source ...
Maximum Render Quality
This is a high-quality resize operation that should be used when outputting to a different frame size from your sequence. It can reduce aliasing (jagged edges) when resizing images but is of no use when outputting to the same frame size. This operation significantly increases render times so only use it when resizing.
Render at Maximum Depth
This renders content at 32-bit color depth. Very few output formats actually support 32-bit color but processing at this depth can produce better quality for compositing and effects operations before being scaled back to the output format's bit depth. It can reduce or eliminate artifacts and banding in your video but that benefit comes at the cost of an increase in processing time, so only use it when completely necessary.
You may benefit from this option in the following situations:
- Your source media has a higher bit depth than the format you are outputting to
- Your sequence contains heavy compositing or lots of layered effects (particularly 32-bit color effects)
- Your sequence contains very high contrast or very low contrast images (for example subtle gradients)
Although I do a modest amount of resizing on export, and at times go a bit nuts in Lumetri, I've not found either of them very useful (as far as improving an export) since about PrPro CC2014.
Multiple-pass exports are also something I've not bothered with since about CC2014. It just doesn't do any better than the first-pass process seems to do these days.
I've tested exports using each of these imported back into the project, and looked at them with the Program monitor set to 200%, and couldn't see any improvements. I'd suggest if you're concerned, do the same tests yourself. Takes maybe half an hour to do a few quick exports & review them.
If something slows the entire processing change down for results that are impossible to quantify, well ... it's only wasted time.
Neil
Thanks Neil... I am generally the sort of person who will do side-by-side tests to compare results for myself but the nuanced info you have above really helps to make better choices along the way despite any personal informal findings. I mean, I could do (and actually have done) side-by-side tests with my usual DSLR source footage and draw conclusions from that, but it wouldn't give me the resizing and 32-bit effects hints (and more) reflected above... nor would it clue me in should I start using source footage having a greater depth. Or... the one that really gets me... despite my tests showing no immediate apparent benefits, I use the settings just in case my tests didn't cover the mysterious cases where there's a benefit... elaboration as you've presented, I'm saying, removes that mystery and allows one to get more money for my own personal testing. Unless of course I test for each project, wasting time each project, to compare that the project's specific cases so to speak... rather than knowing what's under the hood here. Having just a hint at least can help, is all I'm saying. Wish this were clarified release-by-release in a readme or blog or something.
So I'm working on a project right now that has some source footage (not photos but video) that is HD or other smaller odd ratios than 1080p (1080p being the output). Since those are scaled up to 1080p in the main timeline, would you recommend pre-rendering those to a mezzanine using MRQ... then pulling that into the main project... those sections don't need much ongoing work from original source so it seems taking care of MRQ upfront with a pre-render might be a good strategy for those effect "resize" operations... then avoiding the MRQ penalty on the main project. Currently, I've been going to CineForm with those but keeping at the original scaling, then scalling the Cineform in the main project... seems this is a case for maximizing scaling processing before the mezzanine (I tend to avoid over processing at that stage as I've gotten burned by doing too much, trying to be too efficient, before heading to mezzanine). What do you think on that?
Thanks for the 2pass info... that was going to be my follow-up Q. 
... okay, so I'll stay away from that for now. ... I'll also take some time out to do some experiments with my particular footage... but with some of the mystery removed the tests themselves will be more meaningful. Thank you for those clarifications.
https://forums.adobe.com/people/R+Neil+Haugen wrote
That bit you quoted about MRQ ... that often with highly-compressed media it makes things worse? Yup.
Good to know!
https://forums.adobe.com/people/R+Neil+Haugen wrote
The majority of my media starts from either the GH3, which I've typically shot in "standard" mp4 or mov (both highly-compressed long-GOP) as the stuff has more noise when using the All-Intra camera setting; and these days, for some of my projects, from my Samsung S7 ... phone! You know that's highly-compressed long-GOP.
For much of the 'standard' mov/mp4 from the GH3, if the project needed best quality, I've used either MediaEncoder to watch-folder create transcodes, or Prelude to transcode editing media, typically in Cineform, although occasionally DNxHD/R. The phone media, I always convert to CFR via Handbrake, and for some projects, have then taken that 4k/8-bit media and within ME have made 1080 4:2:2 10-bit ... which surprised me as to how far I can push that media then in Lumetri or Resolve without inducing artifacts.
Interesting that "All-Intra" had more noise! Correct me if wrong, but this generally means the in-camera processor/compressor was giving you something ready-made that, for you to solve it with All-Intra back at HQ would have meant transcoding All-Intra to something achieving the same results as GH3 in-camera processing... and in the end you found transcoding the in-camera processed footage (with less retained data) to be viable aesthetically and otherwise.
CFR via Handbrake ... have not heard of doing that with S7 media... I have an S7, have not used handbrake. It seems whatever you transcode to, in that case 1080 4:2:2 10-bit, actually makes use of Lumetri more flexible than on orig footage... so the Lumetri's effectiveness/flexibility seems affected by the source footage for the clip within Premiere, that Premiere doesn't translate during export enough to make that transcoding unnecessary? I was thinking Premiere might have tranlated internally first enough to avoid that... like with MBD chosen... but apparently you see S7->CFR->1080-4:2:2-10bit-ThenUseLumetri a different/better experience. (?) It's interesting because trancoding isn't adding quality (via data) per se... but I guess the space things end up in, perhaps with how the data is worked into that space, allows for better use Lumetri or some such.
https://forums.adobe.com/people/R+Neil+Haugen wrote
Red Giant's video de-noising plugin is doing such an improved job over a version or two back, that I'm starting to shoot more of the All-Intra from the GH3, and just allowing time for de-noising that media. But still, haven't noted that in the 8-bit it produces any advantage to either MBD or MRQ. I haven't tested it for projects needing to go down to 1280x720 or SD yet, though. Especially going to SD for DVD purposes, it might make some difference. I think I can get away with just BluRay finally, however ... so ... well, I'll still test it. Moving diagonal sharply focused lines are marvy for testing this.
Neil
"Red Giant's video de-noising plugin" ... you have an advanced workflow using quite a bit outside of the Adobe suite... I'm currently largely within the boundaries of the suite... still good to hear about these small nuances... even if I don't act to try some of them out right away, they are in recesses of grey matter so can bubble up at convenient times... 
... thank you.
The GH2 with 'hacks' was rather legendary. And the GH3 was coming out with All-Intra, a doubling of the data-rate to file. As quite a number of GH3 users were testing the heck out of it, it was rather ... stunningly ... disappointing that the new built-in feature to give double the data was ... not so good. What apparently happens is that the All-Intra processing of the media for writing in-cam does less pixel blending than the more compressed processing does ... therefore, more actual sensor noise remained.
By the time it processes the media farther to reduce it to half the size, well ... between processing & compression, there's a lot less written noise. Comparing clips shot with All-I and then de-noised compared to a "straight" output ... was a wash. And with some de-noising, you lost enough detail to be less than satisfactory. So at that point, the standard media took less data space on card so you could record more, took less space on the computer, and was as good or better for processing. Probably better exports after post-processing. So, nuts to using the All-I mode.
Now, a couple years later, it seems the Red Giant de-noising is very much improved, especially when you bring spatial in ... where it checks say from 5 frames either way ... to determine what's noise & what's detail. So for me, suddenly ... the All-I media doth give a lot 'thicker' file, and is some better when looked at 100% after post. Not much, but some.
As to using Handbrake with my S7 media, that's all about getting the video from the VFR that the S7 and all other phones shoot into CFR which is what PrPro needs for best work. It does a better job about this than AME. I shoot everything from the S7 in 4k, and mostly, it stays there when I process. But at times I will be putting out 1080, and for that, bringing the 4k CFR output from Handbrake into Adobe's Media Encoder app (that ships with PrPro) for a transcode can be very handy.
Why?
4k media, in 4:2:0 color, can be transcoded to a full 4:2:2 color when down-sized to 1920x1080. Trading resolution bits for color bits, and though some scoff ... it's quite a proven concept both theoretical and empirical. I can "push" things a lot farther and with bigger steps with the S7 t-coded to 1080 4:2:2 than I can with straight S7 4k media. But trying to get there in one step via Handbrake doesn't do it. There's no option to set 4:2:2. And it outputs only 4:2:0.
See the kind of ridiculously nerdy stuff you start to pick up in video post work?
Neil
AdChoices