@petermg

If the processor is at least a 4th generation or later Intel i5 or i7 series then HWA is possible using the Intel GPU.

However, with an AMD processor, then there is no capability to use HWA, unless the GTX 980 is replaced with a model that is supported eg a GTX 5xx series.

For gaming the GTX 980 is extremely fast, however it is a one way process - throw the graphics data at the card for it to render and display.

Video rendering using the gpu for HWA, is a 2 way process, throw the graphics data at the card for it to render and then get the rendered data back and save it disc.

This adds 2 extra steps and introduces other factors which restrict rendering speed, eg hard drive configuration, disc read/write speeds, disk cache size, SATA speed, etc.

I have found that using SSD drives does not provide a significant difference in rendering speed.

By significant, I mean at least a 20 - 25 % increase in speed and preferably higher before investing large sums of money for one with comparable capacity to a hard drive.

HTH

John EB

Hi, welcome to the Magix forums.

There are far more knowledgeable experts than me on this forum but I can quite categorically say that on my computer, running the (slower) AMD GPU than you mentioned, an AMD R7 360, along with the Intel HD 530 Graphics that my i7 CPU offers, I can use HWA on suitable exports without problem.

As you say, it is a known fact that the more recent Nvidia cards have dropped support for CUDA; that was an Nvidia decision, nothing to do with Magix.

Jeff

@petermg

If the processor is at least a 4th generation or later Intel i5 or i7 series then HWA is possible using the Intel GPU.

However, with an AMD processor, then there is no capability to use HWA, unless the GTX 980 is replaced with a model that is supported eg a GTX 5xx series.

For gaming the GTX 980 is extremely fast, however it is a one way process - throw the graphics data at the card for it to render and display.

Video rendering using the gpu for HWA, is a 2 way process, throw the graphics data at the card for it to render and then get the rendered data back and save it disc.

This adds 2 extra steps and introduces other factors which restrict rendering speed, eg hard drive configuration, disc read/write speeds, disk cache size, SATA speed, etc.

I have found that using SSD drives does not provide a significant difference in rendering speed.

By significant, I mean at least a 20 - 25 % increase in speed and preferably higher before investing large sums of money for one with comparable capacity to a hard drive.

HTH

John EB

 

 

 

 

Guys thanks for your posts! Sadly my 6th Gen i7 does NOT have any hardware encoding capabilities. It's a 5820K, got it for $300 on-the-cheap and no Quick Sync or any type of built-in hardware encoding abilities... :\ I got it for it's 6 cores with hyper threading and that I overclock it to 4.5 or 4.6 GHz with water cooling. I know that some people are able to use hardware encoding with older hardware but I mean even the FREE app XMedia Recode can use the hardware GPU encoder NVENC to speed up it's performance significantly. It's a killer program by the way! That said.. maybe I'll just render my files at the fastest setting possible at like 200Mbps then use XMedia Recode to transcode it to a smaller bitrate quickly?... I'll have to look into that I guess. I really tried to get hardware encoding to work, I even tried those tricks with the cuda dll files... no go...

Ok so I just tried that and had "Calculate Effects on the GPU" enabled, which always seems to work although hardware encoding doesn't...??? Anyhow I was upscaling a Full HD video to 4K and noticed everything looked like it had a very small screen on it, so I did a test render of a smaller part of the video without the "Calculate Effects on GPU" enabled and it actually looked better??.. So now I'm re-encoding that entire thing without that option. Guess I don't want effects rendered on my GPU...

Hi

. . . . my 6th Gen i7 . . . . It's a 5820K . . . . I got it for it's 6 cores . . .

It is actually a 5th generation processor, however the lack of integrated GPU does, as you quite rightly say, means no capability for HWA.

I would be interested to know, as the rendering is being done by the processor, how many cores are actually being used.

. . . . I was upscaling a Full HD video to 4K and noticed everything looked like it had a very small screen on it . . . .

Depending on the videos final destination/playback device, it may be better to leave it at Full HD and allow the viewing device, to do the upscaling.

The upscaling is going to make the image appear softer, if you have not done so, try sharpening the output using Effects, Movie Effects, Sharpening tab, Sharpen image weak as a starting point - this will sharpen the entire movie on export or burning to disc.

. . . . Guess I don't want effects rendered on my GPU... . . . .

Correct, this is not required, especially if you were also using HWA, it can cause issues.

HTH

John EB

Hi

. . . . my 6th Gen i7 . . . . It's a 5820K . . . . I got it for it's 6 cores . . .

It is actually a 5th generation processor, however the lack of integrated GPU does, as you quite rightly say, means no capability for HWA.

I would be interested to know, as the rendering is being done by the processor, how many cores are actually being used.

. . . . I was upscaling a Full HD video to 4K and noticed everything looked like it had a very small screen on it . . . .

Depending on the videos final destination/playback device, it may be better to leave it at Full HD and allow the viewing device, to do the upscaling.

The upscaling is going to make the image appear softer, if you have not done so, try sharpening the output using Effects, Movie Effects, Sharpening tab, Sharpen image weak as a starting point - this will sharpen the entire movie on export or burning to disc.

. . . . Guess I don't want effects rendered on my GPU... . . . .

Correct, this is not required, especially if you were also using HWA, it can cause issues.

HTH

John EB

 

Ah, ok 5th gen. Anyhow I upscale it because it's mixed with native 4k footage, that's the only reason, and I am actually sharpening it. It's footage I took with a GoPro Hero 2 stereoscopic set up, with the GoPros set to ProTune, which produces a rather flat and unsharpened image. I'm re-rendering it right now and actually ALL of the cores are being used,

but for some reason none of them are maxed out.

Hi

. . . . it's mixed with native 4k footage . . .

That explains why you are exporting to 4K.

Thank you for the image of the cores in use, when the render is complete can you let me know how long it took and the length of the video exported either by PM or here. I would like to compare the render time using 6 cores compared to 4 cores on my PC.

Thanks again

John EB

Hi

. . . . it's mixed with native 4k footage . . .

That explains why you are exporting to 4K.

Thank you for the image of the cores in use, when the render is complete can you let me know how long it took and the length of the video exported either by PM or here. I would like to compare the render time using 6 cores compared to 4 cores on my PC.

Thanks again

John EB

Just finished. It took 2 hours and 12 minutes, (for about 5 minutes or so of that time I was using XMedia Recode to transcode an mp4 file using NVEnc). Exporting to 4k, sharpening on all the GoPro footage set to maximum, some PIP processing, exporting as AVI, using x264vfw codec, set to almost 100Mbps with preset Ultrafast. The file is 36 minutes 12 seconds and 30.8 GBs. I am planning now on running that through XMedia Recode using NVEnc at a lower bitrate.. somewhere between 25-35Mbps.

I am currently running it through XMedia Recode at 2 pass NVEnc 35Mbps. Says it will take 14.4 minutes a pass. Doing 2 pass just for good measure... then uploading to YouTube.

Hi

Thanks for the info.

John EB

Hi @petermg

Ok so I just tried that and had "Calculate Effects on the GPU" enabled, which always seems to work although hardware encoding doesn't...???

This is because hardware encoding is using a dedicated (Quicksync) encoding block on the CPU's iGPU, or in the case of XMedia Recode on the graphics card chip's separate (Nvenc) hardware block. "Calculating Effects on the GPU" is using the shaders of the GPU to do GPGPU (General Purpose GPU) computing - which is parallel processing of the hardware acceleratable effects, as the program carries out when HW accelerating play back of the timeline.

without the "Calculate Effects on GPU" enabled and it actually looked better??.. So now I'm re-encoding that entire thing without that option. Guess I don't want effects rendered on my GPU...

I would not be shy of using this option, especially in your case where you cannot utilise Quicksync. I have never seen worse quality from using the option and have seen improvements in encoding time in the order of 20-30%, but the gain very much depends on the type and amount of effects in your project and how heavily the CPU is loaded. In some cases it can be slower so test encode a short, effects heavy, section of your timeline.

I'm re-rendering it right now and actually ALL of the cores are being used,but for some reason none of them are maxed out.

That is odd considering you are using a considerable overclock as Task Manager displays 100% at the base clock frequency and the overclock registers as more than 100% (in the Performance Monitor). Is this low usage apparent when previewing a complex timeline with effects?

Maybe try it with one of the effects like sharpening turned off...... just for interest.

Are you using the Software Intel or MC encoder?

Peter

I'm glad petermg brought this up again, because if he didn't, I was going to. (In the hope that someone from Magix reads these posts and takes them to heart.)

Look, MEP has one big Pro - but it's offset by one big Con. Lack of GPU hardware encoding.

I've been using MEP since v11. I'm currently at MEP 2016. Last year I bought the same competitive product as petermg. (I think that, in the past, a moderator has deleted my posts when I mention it. So I won't mention it by name here.)

So, I had a ten minute video I wanted to edit from the Great American Solar Eclipse. Nothing fancy - an initial trim at the beginning and end, fade in/out, bring in background music and speed the video up to fit the 1 minute music clip. The video was from a dash cam (actually, very good video), taken at 1920 x 1080. Oh, a 1 degree rotation because the dash cam wasn't quite level during the Eclipse.

I thought "How long can it take to produce a one minute clip?"

The editing went very fast - maybe about 15 minutes. And that's big Pro of MEP. It was especially easy to see the effects of speed changes in the time line to match my music clip. But it took 2 hours to encode a one minute video! (Quad core AMD running 3.6 GHz.) And the image quality of the output wasn't that good. (That was using tuned up settings in the encoder.)

So then I decided to make the same video using the Competition.

The Competition's UI is a pain to use. Nothing I did in using the Competition was as easy as MEP's editing. Even something simple, like fade in/out, is unintuitive.

Also, the Competition comes from the old DOS/Win 3.1 days of no multi-threading. So even at this late date, all their processes are done in separate windows. So, for example, I was able to change the speed of my video. But I had to do it in a separate window, save that work, go back to the time line to see if the video fit and repeat. Not at all like watching the changes in real time on the time line in MEP.

But once I got the editing done, it took only 15 minutes to produce. (GTX 960.) That was even with the 1 degree rotation, which is a CPU intensive process. And the result is very nice looking.

The bottom line is that it took me about 15 minutes to edit the video using MEP, and 2 hours to encode, for a total time of 2:15. It took about an hour to edit using the Competition, but only 15 minutes to encode. (For a total of 1:15.) While it IS more frustrating to spend more time editing, that is offset by being able to see your results a lot quicker. And now that I've done it in the Competition, I'm getting faster at doing things their way, unintuitve as it is.

When it came time to edit my other video of the Eclipse, taken at 2.7K on a GoPro knockoff Action Cam, I didn't even try MEP. That was a 20 minute video, trimmed down to 13 minutes, with some cuts, crossfades, and titles. Took about an hour to edit. But only 30 minutes to encode.

So even tho I much prefer MEP's UI, I am NOT going to upgrade MEP until Magix supports GPU encoding. (The competition has shown that it can be done.) I don't expect Magix to ever catch up the Competition at this late date. But if Magix supported GPU encoding and got their encode time down by half, that would be acceptable to me.

I know the history behind this. Years ago, before AMD came back on the scene, Magix made the decision to team with Intel and support only Intel hardware encoding. That's their choice. But my next CPU will be an 8-core AMD Ryzen or Zen2. That's my choice. The two choices don't match.

Hopefully Magix will see that it made a mistake and change - for the better. Until it does, it is losing non-Intel customers to the competition.

I'm glad petermg brought this up again, because if he didn't, I was going to. (In the hope that someone from Magix reads these posts and takes them to heart.)

Look, MEP has one big Pro - but it's offset by one big Con. Lack of GPU hardware encoding.

I've been using MEP since v11. I'm currently at MEP 2016. Last year I bought the same competitive product as petermg. (I think that, in the past, a moderator has deleted my posts when I mention it. So I won't mention it by name here.)

So, I had a ten minute video I wanted to edit from the Great American Solar Eclipse. Nothing fancy - an initial trim at the beginning and end, fade in/out, bring in background music and speed the video up to fit the 1 minute music clip. The video was from a dash cam (actually, very good video), taken at 1920 x 1080. Oh, a 1 degree rotation because the dash cam wasn't quite level during the Eclipse.

I thought "How long can it take to produce a one minute clip?"

The editing went very fast - maybe about 15 minutes. And that's big Pro of MEP. It was especially easy to see the effects of speed changes in the time line to match my music clip. But it took 2 hours to encode a one minute video! (Quad core AMD running 3.6 GHz.) And the image quality of the output wasn't that good. (That was using tuned up settings in the encoder.)

So then I decided to make the same video using the Competition.

The Competition's UI is a pain to use. Nothing I did in using the Competition was as easy as MEP's editing. Even something simple, like fade in/out, is unintuitive.

Also, the Competition comes from the old DOS/Win 3.1 days of no multi-threading. So even at this late date, all their processes are done in separate windows. So, for example, I was able to change the speed of my video. But I had to do it in a separate window, save that work, go back to the time line to see if the video fit and repeat. Not at all like watching the changes in real time on the time line in MEP.

But once I got the editing done, it took only 15 minutes to produce. (GTX 960.) That was even with the 1 degree rotation, which is a CPU intensive process. And the result is very nice looking.

The bottom line is that it took me about 15 minutes to edit the video using MEP, and 2 hours to encode, for a total time of 2:15. It took about an hour to edit using the Competition, but only 15 minutes to encode. (For a total of 1:15.) While it IS more frustrating to spend more time editing, that is offset by being able to see your results a lot quicker. And now that I've done it in the Competition, I'm getting faster at doing things their way, unintuitve as it is.

When it came time to edit my other video of the Eclipse, taken at 2.7K on a GoPro knockoff Action Cam, I didn't even try MEP. That was a 20 minute video, trimmed down to 13 minutes, with some cuts, crossfades, and titles. Took about an hour to edit. But only 30 minutes to encode.

So even tho I much prefer MEP's UI, I am NOT going to upgrade MEP until Magix supports GPU encoding. (The competition has shown that it can be done.) I don't expect Magix to ever catch up the Competition at this late date. But if Magix supported GPU encoding and got their encode time down by half, that would be acceptable to me.

I know the history behind this. Years ago, before AMD came back on the scene, Magix made the decision to team with Intel and support only Intel hardware encoding. That's their choice. But my next CPU will be an 8-core AMD Ryzen or Zen2. That's my choice. The two choices don't match.

Hopefully Magix will see that it made a mistake and change - for the better. Until it does, it is losing non-Intel customers to the competition.

Yeah and even Intel customers, since I have an Intel CPU, but it can't do QuickSync. You should try what I did with encoding set to an extremely high bitrate but on Ultrafast or the fastest encoding setting, then just run it through Xmedia Recode, or StaxRip using the GPU encoder. Works VERY well for me. And now that YouTube supports X265 my uploading will be faster too (since I get to encode at lower bitrate with the quality I want retained).

Actually, I snail mailed the CEO of Magix two years ago, telling him how well StaxRip worked w/ my GTX960, and suggested that MEP could leverage that to allow GPU encoding. (I cut & pasted the text and put it in the old forum.)

No response received. Later, I read that Magix had teamed with Intel.

So, yes, you're correct that I could do a 2-step encoding process. But it's easier just to use the Competition. (And giggle over how fast it encodes. Especially when it can use its Special Rendering Technology. Can encode a 1.5 hour TV Movie (with commercials trimmed out) in 2 minutes!)

Hi

Especially when it can use its Special Rendering Technology. Can encode a 1.5 hour TV Movie (with commercials trimmed out) in 2 minutes!)

You couldn't really call that a render - it is really just a file copy and has nothing to do with HW rendering. Magix programs have always been capable of this with the selection of "Smart Copy" in the export settings.

You are correct that they are just copying parts of the video that don't need to be processed. I will have to look for "smart render" on my MEP settings. (Wasn't that only for MPEG-2)?

MPEG4 and MPEG2 - with 2 options.

From the MEP manual:-

Smart Rendering: With Smart Rendering you can considerably reduce the encoding
processing strain for MPEG files. With the production of MPEG files, only those parts
of the movie that were changed in the program (e.g. by video cleaning or effects) are
re-encoded. Please note: The MPEG files contained in the movie must have the same
format, i.e. the bit rates (variable or constant), image resolutions and video formats
must match.
Smart copy (quick, GOP-precise copying): This special mode of Smart Rendering
enables MPEG material to be transferred without having to encode it for the target
medium, thereby increasing the encoding speed manifold. The video material cannot
appear to have been altered in any way, only hard cuts (without fades) are permitted.
These won't be executed precisely to the frame, but rather take place at the next GOP
borders. For this reason, cuts should be set somewhat more generously.