Hey CubeAce,

Thank you so much for your reply and yes, it does help me a lot. I will try to find a machine like yours :)

I will give it a try to get a machine with a 6 cores and see what happens. I don't have a signature like yours because I do not have my dream machine ready yet :)

Thanks so much again.

@LeCanadien

I made one mistake above. I should have said I regularly use 15 gigs of ram for 15 minute projects not 5 minute.

Here is a screenshot of how MEP is using my system.

I would also add get a decent CPU cooler. Run a processor hot and it will throttle back a lot. My system rarely gets above 39c.

Six cores should be plenty.

Excellent. Thank you so much again. Ha! I didn't want to ask about the "5 minutes" thing :)

@LeCanadien

Hi

See my signature for my PC spec - the NVidia graphics card is not necessary for MEP - it cannot use it, and is in my system for a different program.

I edit 4K video without playback issues in MEP with the system I have.

I would add to @CubeAce comment

. . . . Run a processor hot and it will throttle back a lot. . . . . .

when the core temperature reaches about 80 - 85 deg C - with modern processor package design the old issues with expansion/contraction reducing processor life have virtually disappeared - and overclocking is not necessary nor useful for video editing.

HTH

John EB

@johnebaker

Hi John.

You have on several occasions now stated that NVIDIA cards are not used within MEP.

Can you be more specific please as my observations on my system don't support that in either playback or exporting. It is probably true that MEP Premium or below doesn't use the NVIDIA card for encoding and relies on the Intel GPU for Hardware encoding of the exports but are you stating that the NVIDIA cards are not doing anything in playback as well or is actively making things worse? During playback I see the processing bounce back and forth between graphics processors during some effects, when titling is in progress and during some types of cross fades.

During exporting I see the same thing whether I have the export mini viewer on or off.

Could this be down to motherboard differences or even types of source files or effects used in editing?

The black box you see under the Task manager just shows me whether a particular disk is reading or writing at that moment in time by the use of yellow or red, up or down arrows.

If you have any other monitoring processes I could use to try to work out what is actually going on I would be grateful for any ideas or pointers.

Ray.

Hi Ray

In my tests for MEP and VPX using the iGPU or NVidia GPU exporting under the following conditions:

• MEP and VPX are in 1 window on the monitor connected to the UHD 630, no pane of the program on the second monitor connected to the NV GPU
• Video mode set to UHD630
• second monitor only has extended Windows desktop and nothing running on it
• exporting Full HD MP4 25fps
• Same project for all tests - mix of Video 4K video and images and crossfades only
• load measured in GPU-Z and Task Manager

Test 1 exporting to MP4 Full HD 25 fps

1. MEP Graphics setting set to iGPU in windows - ave load iGPU 630 37% : NV 0%
2. MEP Graphics setting set to NV in windows - ave load iGPU 630 35% : NV 0%
3. VPX Graphics setting set to iGPU in windows - ave load iGPU 630 35% : NV 0%
4. VPX Graphics setting set to NV in windows - ave load iGPU 35% : NV 0%

Test 2 exporting HEVC Full HD 25 fps

1. MEP set to iGPU ave iGPU load 64% : NV 0% - quick ~ 72% of real time
2. MEP set to NV ave CPU load 64%, iGPU 0%: NV 0% - slow - 100% real time
3. VPX set to iGPU ave load 64% : NV 0% - quick ~ 70% of real time
4. VPX set to NV - ave 42% load. - also very quick ~ 51% of real time.

Test 3 with MEP and VPX closed.

Switch an active window eg a Video playing or scrolling a web page, the load appears the appropriate graphics card that the monitor displaying it is on, load is <5% with spikes up to 19%.

From the results MEP clearly does not use the NVidia GPU for exporting MP4 or HEVC and VPX only uses it for HEVC.

John EB

@johnebaker

I took that into consideration when taking my screenshots John. I was running MEP on the Intel graphics run monitor with nothing displayed on the second NVIDIA controlled monitor apart from desktop icons. No web browsers or anything active other than when I punched 'print screen' on the keyboard.

I agree the NVIDIA is probably not handling export encoding so why are my screenshots showing activity? Not all my video clips contain effects or titles but there is always some NVIDIA activity.

How are you monitoring performance?

Ray.

@CubeAce

Hi Ray

It is in my last bullet point - load measured in GPU-Z and Task Manager - the only issue between the 2 is a small ~ 2% difference which I averaged out over the render time.

There is some very minor activity <2% at random intervals - status bar refresh? - the only time I see anything more than that is shown below

HTH

John EB

Hi All

My experience with the use of the Nvidia GPU is similar to @CubeAce in the splitting of the load with the Intel HD530, although my main point would be that generally the increased activity on NV comes when the "calculate video effects on the GPU" in the program settings is ticked during export, and when the NV is selected in the Program Settings> Display options during timeline playback (my default). My Magix programs are set to "Use Power Saving GPU" in the Windows advanced graphic settings.

I am surprised you do not see more activity on your NV @johnebaker. I see activity during encoding albeit a few percent even when transcoding unaltered 4K XAVCS to HD. With this footage export is actually faster (0.6x realtime as against 1x realtime) with calculate VFX on GPU unchecked but experience has shown that the VFX on GPU option is quicker with a lot of FX on a large project.

Ray I notice in other threads that you have a tick against Overload Optimisation in the first tab of your program settings which from earlier version help files described this as an option for systems with a weak processor. If I recall correctly this puts more work onto the GPU to assist so may be why you see more loading on your NV. I doubt this setting is beneficial in your case with all the horsepower you have under the hood, and there has plenty of evidence over the forum years that this creates a performance hit in any half decent system.

Some interesting discussion going on in parallel on the German Forum here: https://www.magix.info/us/forum/hardware-anforderung-fr-schnellsten-export--1240398/.

@Scenestealer

Hi Peter.

What a wonderful memory you have.😄

I looked and it's blanked out with a tick in it! Maybe it has to do more with the Graphics card than the processor?

Either way I can't change it at present. When I get time I may take the graphics card out and see but for now I will leave it alone. I may have more time on my hands soon whether I like it or not. Looks like the government here has taken away my ability to work full time or maybe even at all for three months or more so I will be be looking at the implications of that over the weekend. Just had that bit of happy news pop into my inbox. 😟 I'm self employed so I've no idea what to expect yet.

Anyway, I've been doing my own digging around the problem and Can find a possible culprit.

This could be how john has MEP set up.

@johnebaker

Hi John. Is this how you MEP

Rather than this?

Ignore the left hand side of the screen shots, I was experimenting with a lot of different settings.

What I found was with the NVIDIA card activated in that box the NVIDIA card did most of the 3D rendering leaving the Intel GPU free to do more decoding of the video.

But my card is so old and lacking punch it's actually slowing down exports although playback is smoother.

Could you check and see if you can get it activated whether your card makes a difference to either playback or export times please?

Hi Peter.

I will look at the discussion now. Thank you for sharing.

Ray.

On the "Overload Optimization" tick box - I can't remember if I had enabled that when I first installed MEP 2020 Plus or not. (That is, if it defaulted to Enabled.) In any event, I disabled it a few days ago, thinking that my system is no longer under powered. I wonder if Magix should have a "Results" text near the box, from when the program sniffs to see what kind of performance you have. (IIRC, it used to sniff your system in the old days during install.)

Or perhaps a stand alone utility where Magix sniffs your system and then present recommended Program Settings for optimal performance from MEP?

@CubeAce, @Scenestealer

Hi Ray

Ray - it looks like we are talking at cross purposes here - I am referring to exporting, the settings you are looking at are for Preview in arranger if I set this to the RTX 2060 then I do see about 15-17% activity when playing the timeline - I normally leave this set to the iGPU.

Overload optimisation is also turned off.

If I set VPX to use the iGPU in Windows then when exporting HEVC I do see exporting sharing the load between the iGPU and RTX, however no shared activity with other export formats.

MEP does not share the load between GPU's when exporting to HEVC

John EB

@johnebaker

Hi John.

I really am trying to be on the same page as you and agree the NVIDIA card does not seem to help much in doing rendering (and I concede probably does nothing in that department at all) but still I feel there may be some benefits to running a video card within MEP whereas in your general replies you seem to be saying there is nothing to be gained by having one.

While the NVIDIA card does not process video in export it does on my system seem to do an amount of 3D rendering. That in turn seems to allow the Intel GPU to increase it's workload share to export rendering as its 3D rendering amount seems to drop when the NVIDIA card is in play. (Looking at the graphs during exporting.) The increase is very evident when a clip has an effect from Pro Dad's Vita Scene plugin applied (and a few others).

My playback is also smoother, as is scrolling back and forth or moving the cursor across the timeline at a rate of knots.

Export time is slowed marginally compared to not using it.

Last night I tried several runs with different configurations including setting the Intel GPU priority in Windows.

Using High Performance in the Windows Graphics settings for MEP, crashed MEP every time.

Using the System default setting as opposed to Power saving increased export times on a four minute video project by around thirty seconds regardless of how I set up the GPU usage in MEP.

There was also little observable difference to the operational handling of MEP either between System default and Power saving.

Using only the Intel GPU gave fastest rendering times by thirty seconds over the nearest rendering time and fastest project loading, but had less smooth playback and scrolling on my system. The final exported files also looked slightly more grainy and less detailed than when allowing the NVIDIA card to do some of the 3D processing.

Using a mixture of Intel and NVIDIA GPUs in MEP gave the slowest export time being a full two minutes slower than the fastest time (the fastest time was 13 minutes and seven seconds). but gave me a much better handling experience in MEP for scrolling and playback regardless of effects in use, and to my eye, (being a subjective assessment rather than a scientific one) a slightly better looking exported video using the same export settings throughout and that was using MEPs own MP4 wrapper.

I have yet to do a similar HEVC export test.

Ray.

@CubeAce @johnebaker

HI

I am very much with Ray on most of his observations and firmly believe the discreet card is a benefit in playback and encoding in MEP. My difference is that Hardware accelerated Export is completely disabled if I set the Windows setting to anything but the Power Saving iGPU, and Magix tech support have confirmed that is is to be expected. I have not detected any difference in the picture quality of the export but then I have not done the direct comparison. The only thing that might change that that I can think of is whether the 'Use Dithering For Output' is checked in the Program Settings.

My feeling is like Rays, in that the Nvidia, in sharing the the 3D rendering is reducing the load on the iGPU which is encoding and decoding, whether it be rendering for export or playback. The dramatic boost to VPX and HEVC is testimony to this where the decoding is occurring on the NVDEC Hardware chip as opposed to everything happening on the iGPU.

Ray - The screenshot with the Overload Optimisation checked was not the one you have just posted and in fact was not greyed out. Maybe as @pmikep has suggested - Magix has detected your weaker card, which as such would not benefit your other beefy hardware and has disabled the option.....?? Who knows, but there are several members I have seen over the years who have found jerky playback with this enabled.

Peter

@Scenestealer@Scenestealer @pmikep @LeCanadien

Hi Peter.

I think it may be the other way around.The box with the tick in it is not greyed out but may be locked in. Why else would I keep getting readings that it is in use? I have found ways and settings that disable the use of the NVIDIA card during exporting but I will get to that later. I did loads more testing last night after reading through these posts (and the posts on the other forum you kindly supplied) that gave me further ideas.

Basically what it boiled down to was me switching settings on or off within various bits of the program one or two settings at a time while at the same time opening the manual to see what Magix had to say about that setting.

I will skip some of the testing results that didn't seem to improve or worsen the playback or exporting of the same project in each run. I also stuck to exporting in MEPs own MP4 format as I know there is no use of the NVIDIA card within MEP (up to the Premium version and not including VPX) as those programs can't make use of any NVIDIA card for hardware acceleration for export and in fact was not what I was looking for. This post is going to be long enough as it is.

So. To date. My setup has nearly always had the following settings in the Program settings since I upgraded my PC except when I have had the NVIDIA card selected in the Video mode box. It has always given me the fastest export rendering times but not the smoothest playback within MEP at cross fade points (dependent on the type of fade effect in use and whether one clip either side has had additional effects added to it). where it has sometimes stuttered or on a rare occasion made the second clip go blank and lost or corrupted the project settings and sometimes crashed MEP. It has also meant the the lightning symbol has always been on and often all but the Turn effects off selection has been ticked for playback.

I had always used the Standard mode (Direct 3D, hardware acceleration.) but now swapped to the Alternative mode 2 and ticked for no interlacing as I realised none of my video clips use interlacing and I don't make DVD or Blu-Ray discs. The result for playback seemed to open up both the Intel and NVIDIA 3D processing on playback and transitions played back a bit more smoothly but any time the NVIDIA card was able to do anything on my system the export times dropped to at least double that of not using Intel graphics card on its own. I put that down to my card not being new or that fast or have that much inboard ram. It does however just make the minimum requirements to be used in MEP according to the manual and has the latest drivers applied so it was in use for all but one of the tests. The results in playback and in rendering were much the same CPU and GPU wise. Where there were effects applied to a clip Both Intel and NVIDIA cards did a lot of the 3D rendering that seemed to throttle back the Intel's ability to Video Decode.

Playback.

Note the amount of cores in use in the CPU. Much more than if I only have the Intel GPU in use where only core 0 and core 15 seem to do most of the work and the rest of the cores sometimes for brief periods show some small amount of action barely above tick-over.

On export things get more interesting.

This section is where I have applied titles and some effects to the same clip. Those being New Blue Tilt Shift and ProDad Vita Scene. The CPU seems to be using all cores to some degree or other but not excessively so. The Intel GPU in 3D rendering is not that much at this point but eventually does go up a bit later on to about half way up the graph. The NVIDIA card also climbed to around 34% of use. This is where rendering on export slows down to a glacial pace and you can see the Intel GPU hardly doing any Video decoding. Once it gets past any clips with effects added it pick up to a normal rendering pace even on transitions of most types.

Then I got to thinking about Peter's comment about most of my system being quite higher end. So the next step was to look at some of the other settings. As I have the project sitting on an M.2 drive I deselected some background settings.

In playback there is now no juddering during some transitions. Scrolling along the timeline keeps up with no discernible lag.Jumping back and forth and hitting play, it just responds.

Exporting wise it made little difference to the results previous time-wise but did seem to lower CPU use and increase the Intel use of 3D rendering while lowering the NVIDIA card 3D render and there was a slight increase in the Intel Video Decode during the clips that had effects. CPU temperature and usage also lowered and rendering in that part went slightly quicker.

You can see the transition point at which the clip with effects applied stops and the increase in the ability of my Intel GPU to pick up the pace on the Video Decode.

So, what next?

I'm not sure how many people have realised but you can disable GPU hardware decoding of all New Blue Effects, so I tried that. Notice I ticked both boxes and then restarted MEP.

The result in playback was the glitchiest I have ever experienced without crashes but look what happens on exporting during the export of the file while it is exporting the clip with the Tilt shift effect applied.

Now all my CPU cores seem to be doing something all the time. The Intel Video Decode has almost but not quite come to a crawl. I added additional monitoring here for CPU and GPU temperatures etc. My CPU temperature that normally sits around 37c averaged 40c. If you look at the voltage section you can see the NVIDIA card is all but shut down with less than 1 volt being supplied to it.

As soon as that clip has been rendered things change and go back to almost normal.

Rendering times go back to about the same as for any other settings used within MEP only being slightly slower.

The CPU temperature lowers but is still a bit higher than usual but the rendering of the effects clip took 39 minutes alone, making it the longest render of that section by another 19 minutes over any other slow export option. After that section was over the rest of the export took roughly the same time as If I had only used the Intel GPU with no disablement of the GPU usage of the New Blue Tilt Shift effect.

I think this is showing me that any slowdown in any part of the chain of components is slowing down the Intel chips ability to Video Decode. Be it making the CPU do 3D render work where it is most probably slowed down by the ram speeds that are much slower than a graphics card ram (Even an older one) or is being made to 3D render as well as Video Decode where a graphics card is in the chain.

Taking my old graphics card out of the chain speeds exporting up no question but playback and editing experience suffers a lot despite using fast hard drives with plenty of free space.

My conclusion therefore is that at least with less powerful graphics cards it is better to export video without the card in use if you worry about export speed, especially when adding effects or titles. For a more rounded editing experience and smoother playback when as in this project I was using 4K 50fps footage, I think that is where the use of a good graphics card is needed and possibly a better card may help with export times as well if it reduces the 3D rendering done by the Intel GPU during exporting. I also think the speed of the motherboards' ram may be playing a part here as it seems to be another bottleneck in the system. But how much gain there is to be had over the additional cost as it still lags well behind video card ram is hard to say.

I'm ready now to be picked apart 😃

Okay, my 2 cents worth:

MEP 2020 Plus, v. 19.0.2.58

Win10 Home 1909, i3-8100 (UHD 630), Nvidia 1650 Super (Turin Hardware Encoder), 16 GB RAM (2400 MHz)

I have two monitors. My Primary monitor is 1920x1080, and is fed by the Nvidia. My secondary monitor is smaller, not quite 16:9. (1680x1050.) It is fed by the UHD 630. (My BIOS provides for Multi-Monitor, and I have set the Nvidia as the primary driver in the BIOS. My BIOS says that selecting this still allows the UHD to be seen by programs in Windows.)

(In my normal work flow, I set the MEP Monitor window in the UHD driven monitor. But for this test, I am using the MEP standard window layout, with everything on my Primary monitor.)

In Graphic Settings, I have MEP set to Low Power, implying that MEP should be using the Intel UHD 630 for encoding. And in fact, MEP allows me to select Hardware Encoding for encodes, as seen in a screen shot below.

For this post, I am using a 4K video sample file that I downloaded for free from the 'Net, Puppies in a Bath if you want to try it too.

(I never do 4K, but I am assuming that 4K taxes MEP and my resources the most. Although from other testing, that might not be a correct assumption.)

Here is the MediaInfo report on the video file:

FWIW, this is a variable bit rate file. (My understanding is that VBR makes encoding more challenging/longer.)

I put the file on the Timeline. I did not do anything to the video. I am simply encoding it again. (That is, no transitions, no color corrections, etc., which probably opens up a whole new world of benchmarking. and differing results.) I figure that no transitions is a "vanilla" test, the most basic possible.

Even so, I enabled "Calculate effects on GPU," as seen in two screen shots below. (From notes to myself in January, enabling/disabling "Calculate ... on GPU" made no difference in my simple testing.)

For my first test, I set the Video Mode in MEP to point to the Nvidia GPU.

I set the encoding settings to match the settings of the Puppies video, as reported by MediaInfo.

I presume that I did it right when MEP no longer complained about frame rates, aspect ratio, etc.

I started the encode. You can see that "Hardware Encoding" is being called out in the Format Description.

Here is a sample of Task Manager.

It took 7:50 min:sec to encode the 2:30 min:sec video.

Then I changed the Video Mode setting in MEP to point to the UHD.

Start encode.

Sample of Task Manager:

You can see that CPU usage is a bit lower, the UHD usage is much higher, and the NVIDA almost zero.

It took 7:37 min:sec to encode. Here it seems that "using" the UHD is slightly faster. (But see my next post, below, where the Nvidia is faster.)

I did not have "Video Preview" displaying during encode. So I wouldn't expect to see any difference between GPU usage. (Since no GPU would be busy with screen draws.) Nevertheless, you can see a difference in usage.

Another test.

This time, a 1920x1280 30 fps from my Canon camera. (The Great American Solar Eclipse.) 1:09 min:sec clip. Again, no transitions. I simply encoded it to the same specs reported by MediaInfo. (CBR this time.)

Task Manager during encode with the Nvidia selected in Video Mode:

It took 1:10 min:sec. Hardly any activity shown for the UHD.

Now Task Manager with UHD selected:

It took 1:23 min:sec. I'm going to say faster with Nvidia selected.

I don't know why. But it is.

@pmikep

Can you give me a screenshot of what the Intel GPU is doing on both please?

Ray.

Here is a screenshot of the Intel UHD during export of Puppies 4K, with UHD selected for Video Mode.

And a screenshot of Intel UHD during export of 1080p Solar Eclipse, with UHD Video Mode.

@pmikep

Hi Mike.

No, what I really need the the GPU usage under the same conditions as your first two examples.

Look at the differences of GPU usage in your first two examples.

I need to know how much is 3D rendering and how much is Video decoding with the settings you used.

Your second set seems to show different results to your first set.

Okay, I think I know what you're asking for now.

Here are two screen shots of the UHD GPU. The first is encoding the puppies 4K video with the Nvidia set in Display Mode. The second is puppies 4K with UHD set.

@pmikep

Hi

. . . . My Primary monitor is 1920x1080, and is fed by the Nvidia . . . . I set the MEP Monitor window in the UHD driven monitor. But for this test, I am using the MEP standard window layout, with everything on my Primary monitor. . . .

What do the charts look like if MEP is displayed on the UHD driven monitor ?

John EB