@Scenestealer

Hi Peter.

Taking your last point first.

Both were in use when I got the lowest export times. I've no idea why but with one or the other not ticked the export times were slower.

I am now getting even more confused while trying to setup a complete UHD project I could send to everyone as a complete project of under 2Gig via Wetransfer.

I accidentally exported my rendering to my J drive. The one I was reading the project from. This time I got what has turned out to be my best export time 0f 15.01 minutes from the 2minute 21.5 second project. This was with only the calculate VFX on GPU un-ticked but still have Nvidia set for display settings.

I'm re-running that export again now to double check my findings.

Anyway I have made a UHD project with lots of varied effects and types of cross fades as well as colour corrections and animation of lots frames of one or two jpgs. I'm also using 12bit 48kHz wav. files for audio. It's a track I put together so no copyright problems. Anyway, I'm just coming to the end of the export and it's coming in at..15: 47s

I should add, this export is to my D: drive WD black. it is one minute slower than the export I did to the J: M.2 drive.

This is nuts.

I will make the whole project available to anyone that wants it. I assume it will have to be run on the latest version of MEP Premium.

Ray.

@CubeAce@CubeAce

Hi Ray

Can you include a set of test conditions and the results you get in a text file or PDF so we are all singing on the same page when testing.

. . . . From there we could add some agreed effects to each or even apply these to the test project that is uploaded and they can be switched off in the KF timeline for the first baseline encode. . . . .

I would extend Peter's comment to include a set of test conditions to take into account PAL and NTSC users and associated framerates for things like exporting as AVCHD for Blu-Ray

John EB.

@Scenestealer

@Scenestealer

Hi John and Peter. (or anyone else with nothing better to do.)

If I sent the whole project would it not include the settings ie, frame rate frame size etc?

The only things to be lost would be graphic GPU settings and export destinations wouldn't it?

As far as I know I've only used effects from MEPs basic sets but in combinations Clicking a section should bring up what is in use in the Key-frame editor wouldn't it?

I'd been working on the base project for months. It may well have over a hundred different things done to it.

I don't keep a track of such things.

it contains two 3840x269 at 50fps 192kHz dual mono aac files set at present for similar export settings.

The rest of the files are four jpgs of which are copied across parts of the timeline in two groups. Three 24bit 48kHz Wav files.and four MP4 short files of animations I did. The rest are applied effects. There are no pre rendering files or proxies. They can be added by the user if needed. I don't need to render or use proxies when using my M.2 drive to read from. For all I know the pre-render requirements of a different machine may need more or less.

If the project can't be opened as is apart from GPU configuration then there is little point.

All I'm trying to find out is if anyone can export the project at it's original file size in MP4 format (or other after if you wish) and see what export times are like.

I can put some jpgs in another folder chowing settings I've made for the project and exporting if that would help but mainly I would like to see what others export times are like. What makes a difference etc.

Here is what the finished export should look like. It is only visible via this link.

It's available at full 4K but some auto settings will put it at 1080p.

I've been keeping to the parameters of the original files as close as possible for export, same frame rate size etc but keeping to using 'Best' and sticking to MEPs MP4 file container. Only tweaking things like un-ticking or ticking various export options and GPU selections, Internal MEP setup etc to see what gave me the best looking export and the fastest render times.

Total project size should be 32 Files, 1 Folder 1.59 GB (1,713,537,024 bytes)

[EDIT]

Have now added a folder of three jpgs showing all the settings I used for my best export time as well that should come with the project folder.

Ray.

@CubeAce

Hi Ray

. . . . If I sent the whole project would it not include the settings ie, frame rate frame size etc? . . . . The only things to be lost would be graphic GPU settings and export destinations wouldn't it? . . . .

That is correct.

So long as the project is using standard features of MEP eg normal transitions, titles - except the Dynamic ones, keyframe animation etc then there should be no issues.

. . . . There are no pre rendering files or proxies . . . .

I forgot about that one - as you have said if the user needs them then they can create them.

. . . . Have now added a folder of three jpgs showing all the settings I used for my best export time . . .

We will need your export times for comparison with the project files.

For comparison this is the Vegas Pro 4K bench marking project, I have downscaled it to 1920 x 1080.

As you can see it has a lot of colour effects and a slow motion section. It cannot be translated directly to MEP/VPX as it uses features we are not standard in MEP/VPX

John EB

@johnebaker

@Scenestealer

Hi John.

My best export time was 15 minute 45 seconds John under the conditions mentioned above. Have not tried an export to a lower resolution as yet. There is no use of dynamic titling in the project. Type faces I think are Android and Arial. I will check and correct that info here If I'm wrong and add an [Edit] alongside any corrections. All effects and transitions are also standard MEP ones. I will double check now. It may take me a couple of hours.

My best 4k standard MEP MP4 export time was accidentally obtained by both reading and exporting to my M2.1 drive.

Ray.

@CubeAce

Hi Ray

This time I got what has turned out to be my best export time 0f 15.01 minutes from the 2minute 21.5 second project. This was with only the calculate VFX on GPU un-ticked but still have Nvidia set for display settings.

Well that certainly agrees with my findings of late where ticking calculate VFX on GPU consistently slows down the export and setting the Nvidia for Playback in the arranger always accelerates export rendering. Not to say I have not seen faster exports with Calculate VFX on GPU selected on larger 4K > HD exports whereas here I have been doing relatively simple project tests and therein may be the answer ie it may depend on how much the hardware is being stressed as to whether the discreet GPU is helping or hindering?

Thanks for offering your project for testing although I was envisaging something a little less heavy duty than UHD 50P and so many tracks, but in light of my previous sentence I would be happy to give it a go seeing as you have it already "baked".

@johnebaker

I am interested in how they have configured that Vegas Pro 4K test but I can not get the link to open.

Peter

@johnebaker

@Scenestealer

Hi Guys.

Well I've gone through all the effects used. They are all MEPs own as are the fonts used. Not Android as I first tried to remember but Ethnocentric and Arial.

So the list of effects in no particular order or frequency of use are.

Color

The one transition between the two video clips is Effect Transition Basic / Blurry clockwise around central point

Distortion / Whirlpool / Fisheye / Mosaic

Lens / Flip vertical / flip horizontal / kaleidoscope / Glow

Image size position

Reverse Play

Lens correction / section zoom.

Along with the other specs I gave earlier which were

four jpgs of which are copied across parts of the timeline in two groups. Three 24bit 48kHz Wav files.and four MP4 short files of animations I did. The rest are applied effects. There are no pre rendering files or proxies. They can be added by the user if needed.

So that's the complete project and maybe not as much as you would think.

Hi Peter.

At least I will find out what is mightier, your Nvidia card or my Intel GPU although after the export and read write surprise I got from the M2.1 drive I'm not sure about anything any more.

Ray.

@Scenestealer

Hi Peter.

I've taken the liberty of sending you the project via Wetransfer that seems to have tightened it's security protocol since I last used it (no bad thing) to see if it will at least load at your end. It's still uploading and should be ready just about ten minutes after midday GMT. [Edit] It has now sent from my end.

[edit2] John EB also now has the project.

Ray.

@CubeAce

Thanks Ray. It is coming down the pipe now.

@Scenestealer

Hi Peter.

I find it cool I've sent something via the internet into the future. It's still the 3rd of April here. 😎

Some thoughts on benchmark videos: I don't have the patience to wait for a 15 minute decode, especially since MEP doesn't report the Elapsed Time after the video is done. (That is, I walk away from the computer to do something else while it's encoding, then come back too late to see what the ET was showing just before completion. But if I'm too late getting back to the computer, I can't tell how long the encode took and I have to do it again.) While I realize that there would be less variation in a 15 minute encode vs a 5 minute encode, I favor shorter rather than longer.

Some effects are going to be more CPU intensive - or GPU intensive - than others. Although I don't know what ones. E.g., in :Photo Editing tools, we know that Gaussian Blur is the gold standard test of how good you CPU does FP arithmetic.

So I'm suggesting a baseline, easy, not very hardware dependent benchmark (as someone recommend above, with simple cross fades).

And then anther benchmark with, say, a color correction (?) which I believe is CPU intensive.

Then one with some effect that's GPU intensive.

This way, users like Ray might get a sense whether it's worth it for him to invest in a newer GPU, or if he's fast enough with CPU power alone. Etc.

@pmikep

Hi Mike.

What's to stop you from turning off or keeping on individual effects I've applied to the project I've already created?

All you have to do is go to an object. Open the keyframe editor and click on one of the effects listed in the box of the key frame editor window to see how it's applied in real time as it brings up the effect above the keyframe window. Or just un-tick the effect for that object. The project is only 2 minutes 21.5 seconds long.

Ray.

True, there's nothing to stop me from undoing individual effects. I'm just used to the Blender BMW paradigm for benchmarking: Just run the file and report results.

@pmikep

Hi Mike.

Well you are welcome to have the project if you want it.

Ray.

Yep. Sent you a PM asking for the link. (Since I didn't see it here?)

@pmikep

Hi Mike.

Have sent you a PM here with a complete explanation of how the files are transferred and the file is being uploaded as I type. Give it five minutes. I have a reasonable upload speed.

Ray.

@pmikep

Hi Mike,

All done. Have fun.

Ray.

Hmm.... unless things speed up here, MEP is reporting 1 hour 45 min to encode Ray's file here. CPU is loafing along at 38%. GPU's are barely moving above 2%. (It's going faster now. CPU just jumped to 95%. UHD kicked up to 32% in spurts. But now down to 50% CPU.)

@pmikep

Hi Mike.

Yes, you will find the titling section at the beginning with most of the effects applied will be reasonably slow ( it reports a maximum time of 1 hour 52 minutes (max) at that point at my end) and then speeds up during the video sections.

Ray.

Yeah, saw that. First pass with Nvidia as Display Monitor, no VFX on GPU=18:16.

Am now trying with FX on GPU. And will do same two tests using UHD as Monitor.

@CubeAce, @Scenestealer, @pmikep

Hi

The results are in for me, I also tested in Video Pro X setting the GPU in Windows to both graphics setsand these are the results for the Project as sent:

With the following additional notes:

The intro and outro takes 6 - 8 minutes in total to render 24 s of timeline.

CPU was running at 4.28 GHz constantly (turbo boost kicked in) over a base clock speed of 3.19 GHz and showed no signs of click speed reduction in any of the exports.

Apologies in advance Ray.

I took the liberty of exporting what I assume would be the standard parts of the intro and outro to MXV format, ie rendered tracks 4 -10, 12, 14-17 to 3840 x 2160 50fps - I did not include the 2 titles as these may change from project to project.

The combined tracks were removed and the MXV files substituted - then ran the exports again with these results:

Apologies for using images - the forum lost the tables function a while ago so cannot do neat consistent layout.

I did not note the CPU/GPU loadings for the MP4 - neither the CPU nor GPU broke into a sweat over the exports.

@Scenestealer

IIRC you also saw strange results when in VPX with render times being extended when both GPUs appear to be working.

My observations showed that the UHD 630 was video decoding and encoding but the RTX 2060 was encoding only!

I have a suspicion that the load is not being shared, the iGPU is taking the source data, decoding it, encoding it, possibly to an intermediate format or data streaming it to the RTX for final encoding 🙃

John EB

Okay, final results: (Update: All with Win10 Graphic Settings set to the UHD for MEP.)

Best time was with Nvidia set as Display Monitor, no VFX on GPU =18:16

Then same, but with VFX on GPU =22:18.

Slower. So calculating VFX on GPU here made things worse.

Then UHD, no VFX on GPU =28:15.

Much worse than with Nvidia. (So there is some merit in having an external Nvidia GPU. Although how fancy a GPU remains to be seen, as below.)

Then UHD 630, with VFX on GPU =25:20.

Better with VFX on GPU. But still not as good as Nvidia.

So one interesting take away is that it is faster to have VFX calculated on the UHD when the Display Monitor is set to UHD. But slower when VFX calculated on Nvidia when the Display Monitor is set to Nvidia.

Another interesting take away is that Ray is running a processor that runs cores up to 4.7 GHz (unless he overclocked more) to my 3.6 GHz. Just doing a basic calculation, if I take Ray's best time (15 min) and multiply it by the ratio of his CPU speed to mine (4.7/3.6) I get a calculated encode time on my system of 19:35. (Note that my CPU rarely ran more than 90% during some parts of this benchmark. So if I am CPU bound, it was only for a short time.)

I got a little better than that - 18:16.

Now, I have slightly faster RAM than Ray (2400 to his 2100). And I have a better Nvidia card. So maybe that helps me.

But he has more cores, hyperthreading and more L3 cache. (And his is called "Smart Cache," whereas mine is not.) Maybe that helps him.

If we ask him nicely, maybe he would turn hyperthreading off in his BIOS and run the test to see if hyperthreading makes much difference to MEP.

I don't know of a way to disable L3 to isolate that in a test.

There are ways to disable cores - either in some BIOS's or using QuickCPU. But asking him to par them down to 4 is asking too much I think. (Even if we ask nicely.)

I expect that the differences between his results and mine are due to his faster clock speed. Which is what all the Gamers say.

Which, if true, makes me wonder how a Ryzen system would do - except that Magix locked out AMD by deciding to require QS years ago.

(Might be time to rethink that decision, Magix. Especially after seeing here how having an Nvidia along for the ride helps MEP a lot.)

It also makes me think that Ray won't benefit much from upgrading his GPU. I suppose we'll have to wait for JohnEB's results, since he's running a 2060. (Update: I see that he posted ahead of me. Except no results with his 2060 as the Display Monitor in MEP?)

As an aside, there was a smokin' deal on PD18 a week ago. So I bought it. Over the next few months, I might try to recreate Ray's effects in that using his assets. (I'm not a PD wiz, so reproducing Ray's effects might take me a long time.) As you all know, PD has chosen to focus on Nvidia HA. So it will be interesting to see its encode times. (As I write this, it occurs to me that I could remove Ray's more complicated effects and just do a simple A/B test on encoding the video.)

I did not compare Image Quality for any of these tests. I suppose, as time goes on, I'll look at my best time encode, scrutinize the video, and if I see anything that looks terrible, I'll check it against the UHD encode to see if there's any difference.