Speed and quality tests - VP19 versus PD365 on a RTX3080 with NVENC.

dape wrote on 8/7/2022, 6:28 PM

So following VEGAS_EricD's advice from another post on this forum regarding adding VP executable to (my) Windows 11 "Graphics settings" as a desktop app with High Performance i found out that both Vegas Pro 19 and PowerDirector 365 utilize almost all the processing power of my RTX3080 ! I am talking about 90-100% encode !

So, i rendered one of my gaming videos (no fx, no cuts, just full 25 min 1080p60 video) and i had the following durations in similar exporting presets:

PD365 - 3 min 5 sec

VP19 with Magix NVENC - 10 min 30 sec

VP19 with Voukoder NVENC - 7 min 20 sec

And then i installed FFMetrics v1.0 (stable) and added all three exports with the reference unedited file (recorded with quick sync in obs) and here is the result:

Which to my surprise put the Voukoder export on the first place regarding quality because i did another tests few months ago and Magix's nvenc encoder was better quality then.

If anyone cares about the edited file to see the motion - it's here: .

Back to post

Comments

Former user wrote on 8/9/2022, 10:41 PM

@Former user my CPU is underVolted, so seeing a change out of nowhere could mean bigger problems. There is a situation where a particular core will crash, or the more insidious problem where cores don't crash but they don't operate at the correct frequency. Think the term is clock stretching, The effective clocks are much lower than reported clocks. It looks like I"m suffering from this, but I know very little about the subject, so will have to consult a site that specializes in this. My benchmark just then was even worse 8m16s. It's still not a big deal, stability fine, and speed is still adequate

Will be interesting to see the results for your new ram

Reyfox wrote on 8/10/2022, 5:56 AM

@Former user I did not use voukoder since I do not have it. I used Magix AVC/ACC to render.

BTW, what are your computer specs? Makes it much easier looking at the signature. Also, rendering @dape's file should all be using Magix since again, I don't have voukoder.

dape wrote on 8/10/2022, 7:09 AM

You tried @dape 's file and got 10m37s, whereas I got 8m03s(voukoder) which is interesting too. His file is computationally quite complex, using 3 B frames

Hey, i have set for 3 b frames in OBS (was maximum on the quick sync encoder) for more quality. I think..

Also.. @Former user - looking at your vbr bitrate.. i have set 25Mbit for exporting, also both dropdowns below nvenc were set at vbr quality. Also rendered with the last two audio tracks muted so i can match my powerdirector365 similar preset.

dape wrote on 8/10/2022, 7:11 AM

@Former user I may have to restart my computer. @dape and I have very similar computers, he's getting 7.20, you're getting 7.31 and I"m 8.03. I know I was suspicious of cpu frequency of 4.3ghz, normally with encodes it's over 4.4ghz

I am using the 12700k on a MSI Edge z690 with 3200cl16 ddr4. I have set the "autoOC" in BIOS for 4.8 Ghz but during rendering all cores were at a stable 4.5 Ghz.

Former user wrote on 8/10/2022, 7:18 AM

.

Also.. @Former user - looking at your vbr bitrate.. i have set 25Mbit for exporting, also both dropdowns below nvenc were set at vbr quality. Also rendered with the last two audio tracks muted so i can match my powerdirector365 similar preset.

@dape Can you show a full screen image showing render template,

dape wrote on 8/10/2022, 7:54 AM

@Former user

dape wrote on 8/10/2022, 8:08 AM

Just updated the game ready nvidia driver (and some security w11 updates) and redid the rendering test with Magix's:

dape wrote on 8/10/2022, 8:18 AM

And with Voukoder's:

Former user wrote on 8/10/2022, 8:43 AM

@dape Ok, adding all that 'High' stuff adds time, I changed the audio to 320 as in your pic, whether that makes any difference I've no idea, I'm not going to do Voukoder as @Reyfox says he hasn't got it & it seems Voukoder is fairly consistent across all

PS Studio driver is normally recommended on here 🤷‍♂️

Reyfox wrote on 8/11/2022, 8:58 AM

@Former user @dape @Former user. In my previous test, I had Dynamic Ram Preview set to "0" that I used for the @RogerS benchmark test, which I will redo after this post.

What I did was increase the DRP to 2% (653MB) and run the test of @dape's file again. Since in MediaInfo, it was 23Mb/s bitrate, I did set the bitrate for 23Mb/s low and 24Mb/s high in the Render settings for Magix/ACC AMD VCE.

And there was a drastic increase in rendering speed over previous test. The first image is of the actual render happening. The second image is when it's finished rendering.

With "0" DRP, render time was 10:37.07 and average fps 141.60. With 2% DRP render time 7:10.38 and average fps 209.6. That's a huge difference and close to what I think I can get out of the card. With another project, I went from 2% DRP to 6%, and I lost 1fps and render time increased a couple of seconds. So I will leave it at 2% on my computer.

Changing the settings in Render to what @dape posted, I saw performance drop to 46fps, GPU to 13% and CPU to 32%. Not sure on whether this will produce a "better" looking finished video. I have another test I have to do so am aborting.

RogerS wrote on 8/11/2022, 9:25 AM

I don't see the point of variable bitrate where the max and average bitrates are the same. The point is to give it heardroom for when it needs more data- it will use closer to the average overall.

dape wrote on 8/11/2022, 9:45 AM

If you guys don't mind one extra comparision with newly released Pinnacle Studio 26, it rendered the cuda accelerated file in 4 (FOUR) minutes and the (buggy, i think) intel in.. 15 minutes:

And the FFMetrics:

dape wrote on 8/17/2022, 9:02 AM

I was curious about mixing my intel video chip into Vegas' soup and..

vp19edit nvenc nvdec magix 12:30
vp19edit nvenc nvdec voukoder 8:37

vp19edit nvenc qs dec magix 12:44
vp19edit nvenc qs dec voukoder 8:31

Reyfox wrote on 8/17/2022, 3:48 PM

The only way anyone would be able to compare is if they have the same files and software.

I did some testing a while back when PS26 was released. Nothing special, just the same clips on the timeline and rendering them out. VP19 on my computer with the "old" RX480 8GB graphics card, was faster. Not marginally faster, but fast faster. PS26 38seconds. VP19 32 seconds with new graphics card.

Clips 1080HD 60P 15Mb/s total length, 2:04

Rendering/export speed isn't the "end all, be all". I spend far more time editing. That is where Vegas excels.

RogerS wrote on 8/18/2022, 12:55 AM

I was curious about mixing my intel video chip into Vegas' soup and..

vp19edit nvenc nvdec magix 12:30
vp19edit nvenc nvdec voukoder 8:37

vp19edit nvenc qs dec magix 12:44
vp19edit nvenc qs dec voukoder 8:31

For this use case I'd just stick with QSV decoding as the times are close and it has somewhat greater compatibility.

Former user wrote on 8/18/2022, 1:39 AM

@dape if you're interested try the difference between NVENC via MagixAVC and Voukoder at different resolutions, I assume that's still 1080P you're testing, speed differences with 720P, 1080P and 4K for me, with 720P having the largest difference between voukoder and MagixAVC