Thanks for your advise.

Hi

That 6 core would be great for rendering

Apart from the lack of iGPU and the rendering speed that would add...... a german member did some reasonably comprehensive comparison tests using 4 and 6 core set ups in similar machines and concluded that a 6 core CPU at a lower clock rate was not as efficient with MEP as a 4 core at a higher clock rate ie a 6800K has a base frequency of 3.4GHz but a 6700K quadcore has 4.0Ghz.

Thanks for the comparison between 6700K & 6800K. Great that someone did the comparison.

I would have always thought that adding more workers will increase the rendering speed. But may be with more workers may be more congestion in getting other resources, so getting the resources faster with less workers actually is better.

Maybe it has something to do with whether the software is optimised to utilise more than 4 cores?

I have no experience with AMD but it would be hard to imagine a Ryzen with 16 Cores would be 4x faster that a quadcore.

Maybe it has something to do with whether the software is optimised to utilise more than 4 cores?

I have no experience with AMD but it would be hard to imagine a Ryzen with 16 Cores would be 4x faster that a quadcore.

That's not the way things work in the real world. That's like saying a quad core AMD will be twice as fast as a dual core Intel, which generally has never been the case. Different applications have differing levels of parallelism. Unless the application is supremely multi-threaded, chances are you won't even get anywhere near close to a 4x increase.

If the application sits on 1-4 threads for most of its work, the Intel is almost assuredly to be faster than the AMD CPU - likely as high as 10% or more, as Intel still has better IPC and per-core performance compared to Ryzen. This is why AMD puts tons of cores at discount prices in their CPUs - they have been doing this for about a decade. Remember all the "Quad Core" APUs they used to sell in laptops, which ended up barely outperforming Atom CPUs with 50% lower clock speeds - on par with i3 CPUs, at best?

Not having an iGPU is fine. The Op can go with that CPU. They'd just need to find an NLE that supports decode and encode acceleration with Nvidia GPUs - like Pinnacle Studio (and maybe PowerDirector, haven't checked in a while).

A lot of software are using NVENC for encode, but that's unnecessary on 7th gen and later Intel QSV, anyways... Many completely skip out on the NVDEC decoding, causing issues when someone wants a CPU without the iGPU.

Also, Ryzen CPUs do not have iGPUs (unless they are APUs), so going that route would not be much different than getting an Intel CPU without UHD graphics... ...

Not having an iGPU is fine. The Op can go with that CPU. They'd just need to find an NLE that supports decode and encode acceleration with Nvidia GPUs - like Pinnacle Studio (and maybe PowerDirector, haven't checked in a while).

Well not thinking of moving away from MEP so not going to find another NLE that supports NVENC.

Seems silly that changing hardware forces to change software as well, may be an updated version but not change completely the brand.

A lot of software are using NVENC for encode, but that's unnecessary on 7th gen and later Intel QSV, anyways... Many completely skip out on the NVDEC decoding, causing issues when someone wants a CPU without the iGPU

I must be dreaming, I saw MEP running HW acceleration when Nvidia acceleration ticked (with intel untick). 😵

@wongck

Hi

. . . . I must be dreaming, I saw MEP running HW acceleration when Nvidia acceleration ticked (with intel untick). . . . .

Assuming this was when exporting video, if so what format were you exporting to and what were the exact words displayed at the top of the export dialog during export?

Which graphics card do you have? I cannot find you computer spec - can you put it in your signature, it would make it easier for us instead of having to ask or search previous posts.

John EB

@Former user

Hi

Maybe it has something to do with whether the software is optimised to utilise more than 4 cores?

I have no experience with AMD but it would be hard to imagine a Ryzen with 16 Cores would be 4x faster that a quadcore.

"That's not the way things work in the real world. That's like saying a quad core AMD will be twice as fast as a dual core Intel, which generally has never been the case. Different applications have differing levels of parallelism. Unless the application is supremely multi-threaded, chances are you won't even get anywhere near close to a 4x increase."

Not sure if you misunderstood my comment but I am saying the same thing as you. Thanks for the additional info though.

"Not having an iGPU is fine. The Op can go with that CPU. They'd just need to find an NLE that supports decode and encode acceleration with Nvidia GPUs"

Sorry but your comment is inappropriate in the context of the original question, in that this is a forum to discuss Magix software and any replies should be relevant to the use of hardware with Magix NLE's.

For most experienced / power users of VPX / MEP, a lack of Hardware acceleration for export, that is at present only possible with the Intel iGPU, would be a major disadvantage.

Peter

Forum Moderator

Which graphics card do you have? I cannot find you computer spec - can you put it in your signature, it would make it easier for us instead of having to ask or search previous posts.

Hey John, so I populated my signature and while doing it, I looked up the driver version of my GTX960 and saw it was the old 2017 8813 version. LOL 🙃 I must have downgraded it to the version that still does CUDA support, I just can't remember that I did that.... 😳

Hi @wongck

You must be dreaming.......Cuda encoding with or without early drivers is just not possible with MEP and recent (Maxwell, Kepler, or later) based Nvidia cards.

I saw MEP running HW acceleration when Nvidia acceleration ticked (with intel untick).

In what window did you perform this tick?

If you are using the Mainconcept encoder and you tick Cuda in the Advanced>Advanced encoder settings you may see Cuda appear in the encoding process window but it will not actually be using it for HW Accel.

Peter

Yes, I better be dreaming because I could have got the CPU without IGPU and still have hw acceleration.

I only recently saw that my GPU0 and GPU1 graph from Windows, either one or the other had a graph depending on the HW acceleration settings in, I think, the program settings windows.

@wongck

You could see activity on the Nvidia card that is not hardware encoding as such but in fact some parallel processing of effects, etc. that is going on during a software encode.

If you are referring to the Display tab in program settings where you can select either the Intel graphics or the Nvidia graphics then you may see more or less activity on the respective GPU's but this is mainly governing which card is doing the majority of the work during playback of the timeline.

Are you seeing better than realtime encoding durations when you see this activity on the discreet Nvidia GPU?

@wongck

Hi

When export rendering, the top bar of the export dialog will tell you if 'CUDA', is actually being used, if you see nothing then it is not being used.

HTH

John EB

@johnebaker

No CUDA on the export dialog just hardware encoding..... well, some sort of it if any.

@Scenestealer

yes, you are right. This is probably the real reason I am seeing the activities when I put the setting on Nvidia.

This is the first time I am planning to export HD videos, all these years I have been doing DVD format only.

So I am really feeling the slowness here now for HD.... for example the screenshot is using Nvidia on the display tab, and it is probably using Nvidia to decode the photos. The main task window is showing Intel as GPU 0 and it is almost zero% while on the smaller graph, you can see Nvidia as GPU 1 going up to 20+%.

Anyway, you can see that it takes 10x the time to encode the video.... movie is at 4.3 secs while elapsed encoding time is 44 secs. No real-time encoding here.... 😢 more like slow-mo.

Setting to Intel for the display tab, drives Nvidia GPU 1 to almost zero while Intel GPU 0 goes to 20%. However, the encoding time is just as bad....

Looks like really no hardware acceleration whatsoever on my system. 😒

Just to add... that I am using MEP encoder that comes with it, no Mainconcept stuff whatsoever. Just everything out of the box and stock.

Ok, so I read the help file... it says only hardware acceleration on AVCHD.

Changing to AVCHD now takes 2 mins instead of 4 mins, still no where near real-time.

@wongck

"Looks like really no hardware acceleration whatsoever on my system. 😒"

Why don't you try to enable the Intel HD4600 in your system for HW Accelerated export, in the Bios?

Or, if you think you have done this then try -

Opening Windows 10 Settings > Display > Graphics Settings and where it allows you to select a GPU to utilise with a specific program select MEP and the assign the "Power Saving GPU - Intel HD graphics."

This is a new setting since Win10 1803 - Worked for me when HW Accel stopped working on my Intel HD530.

Peter

I tried unplugging my LCD from Nvidia and place it on the mobo DVI video out. Same speed noticed.

 

Why don't you try to enable the Intel HD4600 in your system for HW Accelerated export, in the Bios?

It already enabled, I can use the video out from the mobo and see the Intel graghic control panel when I right click on the desktop.

It already enabled, I can use the video out from the mobo and see the Intel graghic control panel when I right click on the desktop.

Then you need to do the next setting I suggested in the Windows 10 "Graphics Settings" to ensure MEP will use the Intel GPU for HW export. If the other two options there are set to either "Default" or "High Performance - Nvidia GPU" then you will likely find the Intel GPU will not be accessed during HW export.

Yes, thanks for the hint on the "Win10 graphics settings" for Magic video Deluxe (that's what it called on mine). It does seems to improve as the remaining time now hovers between 1 to 2 minutes, instead of the 4 minutes.

Somehow, the "Win10 graphics settings" seems to "move" the "MEP Display tab settings" back to Nvidia card. And when I move that setting back to Intel, the "Win10 graphics settings" becomes default. Seems like they are opposite settings.

Anyway I set the "Win10 graphics settings" to Power Saving and left the "MEP Display tab settings" at Nvidia. When I export, now I see both GPU0 and GPU1 graphs having activity instead of only one GPU having activities as noted in the post above. So that's must be improving something 🙂

Hi

Sounds like you have succeeded in getting the iGPU working with HWA for export. How much speed increase you will see during different encodes of course will depend on the format of the source material be it JPEG, MPEG2 or MPEG4, the effects applied, and the MPEG4 preset you are exporting to.

Somehow, the "Win10 graphics settings" seems to "move" the "MEP Display tab settings" back to Nvidia card. And when I move that setting back to Intel, the "Win10 graphics settings" becomes default. Seems like they are opposite settings.

There is definitely an interaction between those settings but all I can say is - if you change one then just check that it has not changed the other. It seems to depend on which one you set first. I think the Windows setting changes the MEP setting to the same but the MEP setting does not subsequently change the Windows setting.

Anyway I set the "Win10 graphics settings" to Power Saving and left the "MEP Display tab settings" at Nvidia.

That is the way I have mine set and in testing it gives the best export times and the smoothest preview performance.

Peter