@johnebaker@johnebaker

Hi John EB.

It always amazes me how different systems produce differing results. Even on the same system (mine) I get different results mainly depending on the project complexity because the source files are from the same camera using the same resolution and frame rate and the exports are the same file formats and settings being used.

Here is from one project that is not doing much other than simple cross fades and colour correction.

This one has picture within picture, text. and speed alterations within this section.

Both are 4K exports using 4K files. Neither project seems to tax either GPUs. Both are exporting to MP4.

Ray.

@SevCardy @johnebaker

This is a snapshot of a 4K project to MP4 - as you can see the UHD 630 is taking most of the load, the CPU a smaller amount and the RTX 2060 on average almost none.

John - my guess is that you have the Intel GPU selected in the Program Settings> Display>Video Mode for playback in the arranger?

On my setup with the Nvidia selected there, it spreads the encoding tasks between the two GPU's with consequently shorter export times, even though that setting is officially meant to only affect HWA for Preview of the timeline.

Sev - I imagine this is what you mean when you say you select the Nvidia for export hardware encoding? There is no other setting you can select to specify one or the other GPU.

John is correct in his comments about the Movie Settings being irrelevant to the source material being Progressive or Interlaced but I find it annoying that a 1920x1080 25P is not selectable for (Edit) .MTS export. It can be selected with the optional Main Concept encoder installed however.

Peter

@Scenestealer

Hi Peter,

John is correct in his comments about the Movie Settings being irrelevant to the source material being Progressive or Interlaced but I find it annoying that a 1920x1080 25P export is not selectable for MP4 export.

Did you mean for export or for project settings?

I see "MP4 FullHD 1920x1080 25p" on the export presets in MEP2021.

John CB

@Scenestealer

Hi Peter,

Here is what I use - non NVidia video card, just an old AMD card:

And here is what I see for MP4 settings:

I see that you've changed your post from MP4 to MTS - foul!

As you can see, I don't have the p for the ACHD transport stream either.

John CB

@browj2

Hi John

Thanks for the correction (Edited prev. post) - I meant .MTS (AVCHD Transport stream) 25P Export using the Intel encoding Preset which only allows 25i. Further - the Intel .MP4 preset does not allow AC3 DD 5.1 surround which I use a lot as my camera records in 25P with DD 5.1 and I watch all my edits on a flash drive or over my home network on TV and want to retain the same format.

Magix say they only have a 25i option in their default Intel encoder because that is the Blu-Ray standard - but I do not burn Blu-Rays.

John is correct in his comments about the Movie Settings being irrelevant to the source material being Progressive or Interlaced

To clarify - Movie Settings = Project Settings.

I see that you've changed your post from MP4 to MTS - foul!

Not!! You are too quick by half John. Actually I straight away re-posted with this post (more or less) and somehow lost it.

Okay, tnx. The interfaces of both look very similar to me. I expect that Handbrake has made some improvements in that department. (Their next iteration will require something called Net 5 (not to be confused with NETFramework).)

In any event, I am seeing faster encode times with Vidcoder. (350 fps vs 280.)

Okay, this is not a Vidcoder (or Handbrake) forum. But just for completeness, I updated to the latest Nightly Build of Handbrake and now it is just as fast as Vidcoder. (Handbrake fixed the "bug" with their slider.) Moreover, the beta version of Handbrake will support isolated processes - which, if I am understanding correctly, allows one to encode to video streams in parallel. (For batch conversions.)

And also, IIRC, the GTX-1050 Super can process 2 streams in parallel.

@Scenestealer

Hi Peteer

. . . .John - my guess is that you have the Intel GPU selected in the Program Settings> Display>Video Mode . . . .

That is correct - my video tend to be mixes of video and images and using the RTX over the UHD does not make that significant, IMO, a difference over the exporting times - eg the project the above images were taken from is approx 2/3 video, 1/3 images, has the following effects applied colour correction, Mercalli Pro v5 stabilisation on some videos, titles, 3 collages, music track and sound effect audio and mainly crossfade transitions and looks like this on the timeline

The project is 8:24 long and export times are

UHD630 display mode - 11.00

RTX2060 display mode - 10:15

As you can see from the images below the RTX does get used for decoding however never gets involved in encoding - both were grabbed at the same timeline time during export.

UHD 630 display mode

RTX 2060 display mode

@CubeAce

Hi Ray

. . . . It always amazes me how different systems produce differing results. . . . .

Surprisingly that is very common, even installing Windows on identical PCs (think 100 - 150 ata time) you can guarantee that there will be installations where the performance is different due to different internal Windows performance settings.

As I have mentioned many times before, what is on the timeline has a significant effect of performance. In the case of the project above, when using the RTX display mode, during export you can see the difference in GPU activity when rendering a video section of the timeline compared to the adjacent block of images - you can also see the CPU spike when encoding the images.

This why we really need a test project like Vegas users have for performance analysis over different systems.

It would have to be a 'real' project possibly one 4K UHD based the other FullHD 1920 x 1080, both h.264 encoded, having a combination of camera footage, jpg images, mp3 audio, standard transitions, effects eg colour correction, resizing etc, titles, basic collages and PiP - no 3rd party plugins or features not available in earlier versions as far back as MEP 2015 if possible.

I did look at the Demo project that comes with MEP/VPX however this is not a 'real life' situation the source video is low resolutuion.

John EB

@johnebaker @browj2 @Scenestealer @pmikep

Hi John.

Mmmm. I wonder sometimes if the Magix increase in performance test results are based on using the supplied demo project 😂 Anyone trying that out on the trial version would think their system could cope with anything.

So we need a "The quick brown fox jumps over the lazy dog" project solution that is not too large a project file size placed somewhere where anyone can access it.

I think it would break most systems at 4K. Plus I'm not sure what a real world average project would look like.

We are seeing people using MEP and VPX for editing game play, video conference calls, phone video output, downloaded content at varying resolutions, stills slide shows, transferring analogue footage, and possibly other things I haven't though of.

It often feels that those of of using the program for video camera footage are in a minority.

Ray.

@CubeAce

Hi Ray,

I agree that we need one or two template projects with video material at UHD and 4K, some photos, titles, music, a couple of transitions and effects, using normal source files, not problematic ones. They could then be tested with Project settings and export at both resolutions. Anyone want to put one together?

I never export to 4K as I have very little footage at that resolution and the only person that I know who has a 4K screen is my mother-in-law, and she wouldn't know the difference.

It often feels that those of of using the program for video camera footage are in a minority.

It could be because those using game play, video conference calls, downloaded content at varying resolutions, etc. are the ones having the most problems. Cell phone video recording is now the norm for most people, if they would just understand that they have to turn the phone sideways and film in landscape. Hmmm, maybe cell phone makers should make landscape the default no matter which way the phone is turned, and warning people about the result of changing to portrait mode.

John CB

@browj2@browj2 @johnebaker @pmikep

Hi John CB.

No, I do really feel we are in a minority. How many people do you think actually own any type of camcorder compared to a phone that use MEP or VPX? How many camcorders use variable frame rates? They want an external editor because most phone video editing programs are complete and utter ****!.

Is there a gameplay editor that allows separate video pip editing? Why do they come to programs like VPX?

It may not be long before there is a TV channel that broadcasts in portrait mode only to cater for the phone generation 😆 with the need for the TV to pivot 90 degrees 👍. (Only joking but many a truth told in jest)

You don't necessarily have to have a 4K screen to notice differences in picture quality even when the 4K footage is shown on an HD screen. Try watching the same clip on YouTube at different resolutions.

TikTok and Instagram prefer portrait video. Recent surveys suggest up to 94% of phone users always use their phones in portrait mode which was a compete shock to me. Older users I think have to accept our 'best practices' may no longer be applicable to the most of the general public.

One of the very first videos I did was off circus acts which I took in portrait mode without thinking of the consequences and a long time before such footage was easy to turn 90 degrees to watch back on a pc monitor.

https://flic.kr/p/8eU5PS

You can see how old that was by the resolution which at the time filled my monitor screen. This is what worries me about working at HD resolution now as that footage was taken in 2010.

What is your idea of a non problematic file? Magix is advertising MEP and VPX are capable of 8K editing and using 10bit H265 files. I'm dreading the day we see one of those queries combining the two.

I know we are only talking at present about testing up to 4K but there are a few YouTubers now using 16K cameras and above. Not to produce footage at such resolution but have the ability to crop without loss of resolution at 4K or 8K. I doubt we will see any of those people. At present the sweet spot appears to be growing for 2.7K for online use. I can record at that resolution but haven't so far tried. I lot of people using action cams are using it though and is slowly spreading for normal use. File size is good and looks slightly better viewed on HD screens.

It is very difficult to keep up or predict trends but they are diversifying quickly.

@CubeAce

Hi Ray,

I did note in my post that most people are now filming with cell phones. The quality is usually very good, depending on the maker of the phone. My camcorder needs a new screen and I haven't used it for several years. I now use my GoPro 5 Hero, occasionally my Canon DSLR, my LGG4 cell phone and my BB. I don't really miss the camcorder.

Problematic files are those that video programs reject, which seems to be those with wildly varying frame rates. That said, I just looked at the details of some of my cell phone videos and those from my wife - Blackberry Z10, LGG4, iPhones. They all have variable frame rates with the BB giving me the largest variation, from 27.65 to 75.591 (recorded in 2014). They all have worked fine in all versions of MEP and VPX that I have used for the last 5 or 6 years. I consider my old BB videos to be the test files - if a video program can't handle them, then it is the video program, not the video file.

I have asked portrait cell phone people if they have also turned their televisions to be in portrait mode. To go to the absurd, there have been a couple of users complaining that they want to see the full height of their portrait video full width on their landscape television with no distortion and no cropping.

Based on what you said, I just switched my GoPro to 2.7K to allow for some zooming in with my UHD projects, especially the cropping/zooming done by Mercalli.

Ok, back to hauling in the firewood from the front to the back of the house.

John CB

@browj2

Hi John.

I don't necessarily believe variable frame rates should be a problem either but each manufacturer seems to adapt the code and change it for their own products and then not share the alterations. How many variations of a codec does there need to be? How many would Magix have to incorporate to get complete compatibility? Worse, how long would old codecs be kept within the program as newer and more efficient ones come out? What would be a good archival codec be for future use?

Digital to me is not like film. the pace of change is much greater. My old 2MP stills looked great on an SVGA monitor. Not so now on HD.

Taken in 2001 with an Olympus 2MP camera.

How long before you can't buy an HD TV or monitor? or LED? Who know?

We are already at 40MP + sensors from 2MP in1998. I expect that to do the same again within the next ten years. I know this has gone off topic sideways now but my point is development of product is always ahead of the software and I think those gaps will only increase over time. It's getting too diverse too quickly to reasonably expect developers to keep up.

Ray.

@johnebaker

I only said the Nvidia is not used for encoding, because you have a UHD 630 this is being used for encoding subject to which parts cannot be accelerated.

Sorry, I wasn't very clear with my explanation. As @Scenestealer correctly surmised, what I meant was: if I select the Nvidia for preview rendering in the program settings, then MEV also uses it for encoding (and I agree the product description does not mention this). When I did my timing tests I verified this by checking the Performance in Task Manager and the Nvidia was definitely being used then (as well as the HD630). Of course, it is not used at all when preview rendering is set to the HD630 in the program settings.

I appreciate that 1080p/25 is non-standard for Blu Ray, and if I were doing this to produce Blu Rays commercially I would certainly have to stick to the standard. But I'm only creating videos for family use as a hobby and to re-live our holidays when I'm in my dotage. Blu Ray 1080p/25 was available as an output option in previous releases of MEP and the disks burnt in that format play correctly on my 4-year-old Samsung BD player and my 11-year-old Panasonic BD player. That's good enough for me so I'll probably continue to use that format until I switch to 4K and HEVC.

What really confuses me is why it should take longer to encode 1080i source material to 1080i H.264 than to a 1080p H.264, which presumably involves deinterlacing. It seems counter-intuitive. Or am I missing something obvious?

Sev.

@CubeAce

Having the hardware acceleration enabled during exporting gives the majority of 3D rendering over to the GPUs leaving more processing available to the CPU. It you are noticing glitches this could be down to one of the GPUs and maybe checking its driver is up to date.

Which 1060 card do you have? There are several variants depending on manufacturer

The drivers are up-to-date (or at least were at the time of the tests - they seem to get updated very frequently!) because I downloaded the latest drivers before testing.

I've an older version 1060 with only 3GB memory and it has always given some problems with MEP. It may be that there is a hardware problem.

Sev.

@SevCardy

Hi Sev.

That could well be the card then. The amount of vram in use on my card on any one project often exceeds 3GB but I'm using 4K content.

Ray.

@SevCardy

Hi

. . . . . When I did my timing tests I verified this by checking the Performance in Task Manager and the Nvidia was definitely being used then (as well as the HD630).  . . . . .

The question is what is it being used for ?

If you look at the 4th image I posted in response to Rays question, you can see that the RTX 2060 is being used, however not for encoding - it is decoding the source video and then passing on the decoded video to the Intel GPU or CPU for encoding into the exported video.

John EB

@johnebaker

Hi John

As you can see from the images below the RTX does get used for decoding however never gets involved in encoding

Unfortunately I can't as the graphs in your images both show the Intel activity and the small overviews at left do not seem to be detecting any activity on the 2060.

Of course to me decoding is part of encoding as the frames need to be decoded before the effects can be applied and then encoded, but that is just semantics.

Why in your 3rd SS is there quite a bit of activity on the 2060 compared to the earlier SS's?

@SevCardy

Hi Sev - interesting that you share my observations on the spreading of tasks during encoding (export).

It surprises me too that 25i to 25i is slower to export than to 25P.

Peter

@Scenestealer

Hi Peter

. . . . at left do not seem to be detecting any activity on the 2060. . . . .

The 2nd and 3rd image - those showing the Intel GPU doing all the work - the 2060 is doing nothing

. . . . Why in your 3rd SS is there quite a bit of activity on the 2060 compared to the earlier SS's? . . . .

That was when I switched the Video mode to use the 2060, the activity in the Video decoding graph, I annotated to show that the video was being decoded by the GPU, however the block of images immediately after were not as would be expected.

The 3D activity, I think, must be the Mercalli v5 stabilisation that was added to the video clips in that portion of the timeline - that was the only effect on them.

I normally keep the MEP Video mode set to Intel, the difference in the exporting times IMHO does not warrant keeping switching the video mode and the 2060 does get used where appropriate.

John EB

@Scenestealer @SevCardy @browj2 @johnebaker @pmikep

Hi all.

Has no-one noticed that even in sections where the nvidia card appears to be doing very little, the amount of the cards' vram in use is often quite high?

Ray.

@CubeAce @johnebaker @SevCardy @browj2 @pmikep

Hi All

I think we need to clarify some things here - not helped by my mis numbering of JohnEB's Screen shots. I overlooked that there were 4 SS's and was referring to the last 3 (of the Task Manager).

Has no-one noticed that even in sections where the nvidia card appears to be doing very little, the amount of the cards' vram in use is often quite high?

Ray - I presume you are referring to the second TM shot where the little overview on the left shows almost no activity but the VRam is sitting at 3 GB? The point I tried to make earlier was that this shot, which again i presume JohnEB was referring to as being a reflection of what was happening on the Nvidia 2060, is in fact showing detailed information on the right reflecting what was happening on the HD630! So we can not see there an indication of the amount of RAM being used on the Nvidia card during that encode.

The 3rd TM shot does show significant activity on the Nvidia but the graphs are correctly displaying for that card and reflect 3GB of VRAM usage which I would expect.

JohnEB - I would not expect to see any activity on the Encode graph in the 3rd TM shot as this only becomes active when the NVENC chip is utilised by the Video Engine during HEVC encoding in VPX11 / 12. MEP 2021 is not capable of using this during encoding.

The encoding load is generally reflected in the 3D graph when not using NVENC.

I believe the CPU spike is in fact being caused by Mercalli which does all it's processing there for preview and export. If you draw a line down to the GPU1 overview the CPU spike corresponds to a slump in the Nvidia decoding overview.

Hi Peter

Great clarification you put what I was trying to say in somewhat clearer terms.

Regarding the 3D graph, do you think the activity there is processing being done on the CUDA cores.

John EB

@johnebaker

Thanks John. Yes I believe it is the Stream processors aka shader cores or Cuda if Nvidia, using Direct X 3D acceleration. I am not certain however how the QuikSync encoding component is reflected in Task manager.

Peter