Thanks @RogerS,

Here's a media info sample:

General
Complete name                            : F:\JamBam\Dodge\20210211-Bargo\Video\2021-08-10\Camera2\P1590500.MP4
Format                                   : MPEG-4
Format profile                           : Base Media / Version 2
Codec ID                                 : mp42 (mp42/avc1)
File size                                : 142 MiB
Duration                                 : 12 s 513 ms
Overall bit rate mode                    : Variable
Overall bit rate                         : 94.9 Mb/s
Encoded date                             : UTC 2021-08-10 17:22:47
Tagged date                              : UTC 2021-08-10 17:22:47

Video
ID                                       : 1
Format                                   : AVC
Format/Info                              : Advanced Video Codec
Format profile                           : High@L5.1
Format settings                          : CABAC / 2 Ref Frames
Format settings, CABAC                   : Yes
Format settings, Reference frames        : 2 frames
Format settings, GOP                     : M=3, N=8
Codec ID                                 : avc1
Codec ID/Info                            : Advanced Video Coding
Duration                                 : 12 s 513 ms
Bit rate mode                            : Variable
Bit rate                                 : 94.6 Mb/s
Maximum bit rate                         : 106 Mb/s
Width                                    : 3 840 pixels
Height                                   : 2 160 pixels
Display aspect ratio                     : 16:9
Frame rate mode                          : Constant
Frame rate                               : 29.970 (30000/1001) FPS
Color space                              : YUV
Chroma subsampling                       : 4:2:0
Bit depth                                : 8 bits
Scan type                                : Progressive
Bits/(Pixel*Frame)                       : 0.381
Stream size                              : 141 MiB (100%)
Encoded date                             : UTC 2021-08-10 17:22:47
Tagged date                              : UTC 2021-08-10 17:22:47
Color range                              : Full
Color primaries                          : BT.709
Transfer characteristics                 : BT.709
Matrix coefficients                      : BT.709
Codec configuration box                  : avcC

Audio
ID                                       : 2
Format                                   : AAC LC
Format/Info                              : Advanced Audio Codec Low Complexity
Codec ID                                 : mp4a-40-2
Duration                                 : 12 s 513 ms
Source duration                          : 12 s 565 ms
Bit rate mode                            : Constant
Bit rate                                 : 128 kb/s
Channel(s)                               : 2 channels
Channel layout                           : L R
Sampling rate                            : 48.0 kHz
Frame rate                               : 46.875 FPS (1024 SPF)
Compression mode                         : Lossy
Stream size                              : 191 KiB (0%)
Source stream size                       : 191 KiB (0%)
Encoded date                             : UTC 2021-08-10 17:22:47
Tagged date                              : UTC 2021-08-10 17:22:47

I've attached a screenshot of the video playing on windows with the task manager.

This is the render performance I'm getting:

Using this standard render template (edit):

This is what the render performance looks like with task manager:

The last sample was a little unfair, but you get the gist. It is a more complicated than average composite as I'm using nested timelines as well but even so the rendering performance is pretty much on-par with my old card using the same settings.

Here's an example of the PIP artifacts in the rendered result, you can see black lines that show the PIP outlines of the blue text, this is new with this card. If I turn off GPU acceleration they disappear, as does even more of my life:

Thanks for your help, I'm hoping it's just something stupid I've missed.

Here are a few initial comments- the file looks like it's constant framerate 4:2:0 8-bit AVC, which should play back using hardware decoding in Vegas.

The task manager view doesn't show much CPU or GPU activity, though. Can you put it on the timeline without any Fx or other events or text? Go to GPU 0 and see if one of the options like copy can be changed to decode? Then play it back and see if there is any decoding activity. If not I'm confused as to why.

For render, is there a way to change the task manager categories to encode? I have an NVIDIA card so the categories may be different, but hopefully there's something like that. Maybe just try a simple render of just the video itself with VCE and see if you get any better times and greater GPU usage. It appears quite low now.

The nested projects may be a bottleneck themselves with render performance (limits to how VEGAS works vs your CPU or GPU). So a new GPU may not solve that.

If you really want to compare apples to apples, try this 4K benchmarking project and do renders with MagicAVC/ VCE and Mainconcept. You can contribute your results and be added to the chart. If you tell me the times for HD and UHD we can see if your system is performing much worse than expected.
https://www.vegascreativesoftware.info/us/forum/benchmarking-results-continued--118503/

Thanks @RogerS,

Looks like that project data has been removed - the dropbox link is empty. I'd like to have a crack though just to compare apples to apples, so if you have another link that would be great!

I'll have another crack tomorrow with simplified rendering while showing the endoding/decoding GPU metrics in task manager, I was hoping I had just missed a checkbox in the configuration somewhere...

Well that's weird on the benchmarking project! I just sent a message to ask. If it doesn't come back soon I'll make a link for you. If you don't see it tomorrow, please tag me again.

Your settings look fine but usage is unexpectedly low.

Using this standard render template (edit):

With such modest source and output, I doubt that you will see any improvement in "performance", whether that means timeline or rendering, with any combination of hardware.

There is no performance increase past "optimal." It takes what it needs, not what you imagine you want.

You can throw a bag of leaves in the back of your $75,000 pickup and it will still "perform" just the same. Give it a 3,000 lb. payload, and you have a whole different story. That's why they test truck engines (and computer processors) under load, not at idle.

Give your CPU and GPU something to chew on, like HEVC 4K 10 bit 60p, and compare "performance." Right now all you are testing is the internal filters, not engine "performance" by anyone's definition.

@JamBam I put together a zip while back that includes the original Sample Project... I see it's still there:

https://drive.google.com/file/d/1dUdyDFehlCSG6Zn7_JSF7NOB0ylxwsC_/view?usp=sharing

This benchmark runs in around 35 seconds on my 11900k and 9900k systems with a 5700xt with little difference between Vegas 18 and 19.

Btw, there were pip fx issues for a while with the 5700xt till the Vegas team fixed em. My work-around at the time was to use a bezier mask that matched the pip position and scale to mask off the problem. Curious how you make out because I'm thinking about a 6800xt myself.

Thanks @RogerS, the drop-box link is still saying deleted. But thanks @Howard-Vigorita I was able to get a copy from your link. I've run a few renders, results below:

Starting with the loopback framerate test:

Yeah, 3.4 FPS :-(

Render 1 - MAGIX AVC/AAC (AMD VCE)

Render 2 - MAGIX AVC/AAC (Mainconcept AVC)

Render 3 - MAGIX HEVC/AAC (AMD/VCE)

Render 4 - MAGIX HEVC/ACC (Intel QSV)

So I think I've established that something isn't right here, the CPU and GPU are on holidays while the framerate is consistently ridiculously low, you'd have to agree @Musicvid these numbers don't pass the pub test. The GPU doesn't seem to engage at all. The video preferences indicate that it is using gfx1032 which is supposed to be the correct RDNA2 APIs:

Thanks all for your interest, I'm open to any suggestions.

 The GPU doesn't seem to engage at all. 

I agree, that seems to be the case. Please post your solution when you find it.

For the first two tests it would be good to compete the render and get proper times. Estimates look like 3:00 and 3:36 or so for VCE and Mainconcept in UHD respectively.

HEVC is ~5:00 and 7:30 for VCE and QSV respectively.

Windows performance isn't showing any GPU decode or encode activity. Is Windows properly reporting this new card? Any reports related to that on the web?

Looking at the results to date there are no results for this card as of yet, which isn't a surprise as it just came out.


There is a Ryzen 71800x with Radeon 7 getting 1:30 for UHD with VCE for comparison.
Maybe you can find other CPU/GPU combinations that are roughly comparable.
My own 2017 era laptop is at 2:48 for UHD NVENC and 3.8fps...

So no, it's not performing as expected.

Even forgetting Vegas for a second, can you download any kind of GPU benchmarking program and just see if it's working generally?

One question-did you reset Vegas after installing the new card?

It might help make sure it's properly recognized and working. You can do this by holding control + shift while double-clicking the desktop (not taskbar) icon. When prompted to clear the cache say yes. You will have to restore customized menus or settings but won't lose any projects or files.

@JamBam I share your disappointment. Some more general purpose free benches I recommend are Cinebench R23 and Novabench. Btw, that board is probably newer than the AMD Enterprise driver so you'd probably be better off with their game driver for a while. Also, if you had a different AMD board in your computer and swapped it out, it's highly recommended that you use DDU to fully remove the old drivers so you can do a totally clean install.

So I think I've established that something isn't right here, the CPU and GPU are on holidays while the framerate is consistently ridiculously low, you'd have to agree @Musicvid these numbers don't pass the pub test. The GPU doesn't seem to engage at all. The video preferences indicate that it is using gfx1032 which is supposed to be the correct RDNA2 APIs:

Thanks all for your interest, I'm open to any suggestions.

Try using the vegas denoise filter and pressing play, On Nvidia the cuda box activity goes up to 100%, On AMD it will be a different box that you need to find see if you get same activity with it peaking at 100%

Thanks again everyone. @RogerS I reset Vegas using your instructions and I think it is slightly better. Here is the full render result using MAGIX AVC/AAC (AMD VCE):

So not a mic-drop moment but yeah slightly better.

@Howard-Vigorita I'm pretty confident the driver is clean, I had to remove the old driver as it was not compatible anyway. Windows recognises it and it scrubs really well in the windows player and utilizes the GPU. @todd-b I tried the denoise but that didn't seem to hit the GPU either.

I'm coming to the conclusion that this card is not supported in Vegas 18, and since 19 is out I'm probably boned.

Still open to suggestions, and I'll update the post if I come across a solution.

3 minutes to 2 minutes is a pretty big jump. Was there any sign of decoding or encoding activity? What about when you just put an AVC clip on the timeline and playback?

You may not be boned- wait until the first Enterprise driver supports your card and test again. It may improve support.

@RogerS Maybe they've implemented machine learning, I've run it again and it's a bit faster again:

I added the AVC clip we started with and it plays back no problem at the full project speed (25fps) and there is activity in the GPU:

I'll re-try my original project and see how it goes.

Perhaps restart the computer if you haven't already. That looks like a good improvement and as drivers mature it may get even better.

@JamBam Just noticed you have qsv templates showing up in the Magix hevc list but not the avc list. As your Ryzen system does not have an Intel igpu, the hevc qsv template must be using a software runtime. You sure none of those legacy boxes are checked? Look at I/O and Deprecated features. Might also uncheck anything Intel in general preferences and take a look in the internal preferences searching for Intel and/or legacy.

@Howard-Vigorita Thanks for noticing. Vegas was set back to factory default (thanks RogerS) which I think slightly improved things, and the HVEC legacy driver is checked after this. This doesn't seem to make any difference to the QSV option appearing even when unchecked. I don't have an intel card, the on-board card is disabled in BIOS so it is invisible to Windows and applications and only the AMD is showing up in the GPU acceleration option under the Video tab in preferences. Searching for "intel" in the system internal settings only shows "Enable Legacy (Mainconcept AVC/AAC and Intel..." which is set to the devault value of FALSE. Searching "legacy" shows the same setting. I think it's as good as its going to get until they release updated drivers. I tried installing the pro drivers (W6600) but they don't recognise this consumer card even though it's based on the same tech.

BTW your bezier masking trick worked nicely to remove the PIP artifacts - thanks! Annoying extra step though...

I think you're getting reasonably decent gpu performance in the Sample Project benchmark, though I expected it would be at least equal that of the 5700xt I had in my xeon system for quite a while. Perhaps a 6700 or 6800 would be a step up but prices are still kind of high for them. But starting to soften. I expect a Vegas update will fix the pip issue as it's very similar to the issue they fixed last year with the 5700xt. Thanks for confirming that there's a work around.

A friend of mine bought a used RX6600 graphics card. Since he has a small case, it didn't fit so he asked me to see if it works. The only thing that was changed was the drivers from RX480 to RX6600.

I used a short clip to test and see if there was going to be a difference. Exporting a DJI Mini 2.7k clip on a UHD 4K time. The render was set to MAGIX AVC/AAC MP4>Internet UHD 2160P 50fps (AMD VCE). I tested both cards. First, the RX6600, uninstall/reinstall RX480 driver (both cards with the same driver revision 22.5.1) and test it with the same clip, settings, etc..

First image is the result of the RX6600. Second image is the RX480. There was a 3X difference in rendering speed. I tested this several times to be sure, and it was. The RX6700XT will be installed in my computer next week.

It would be great to get such a comparison with the standardized benchmark project where only the GPU changes: https://docs.google.com/forms/d/1Exbi4K3hbxw6snJuisR1ble-0tCPVNcIcNnx0BAtSIM/

+1... oh and forgot to include the testing was done with VP19 B643.