Sounds like your content has a lot of still images?

In this case the encoder only writes bits to the first frame and the rest is pretty much just instructions until the next change in content. It uses only what it needs to get the job done.

There are just a few still images near the beginning. The total video is nearly 1 hour and 40 minutes in length. The still images account for less than 1 minute. I'm not sure why it would render the file at less than half of the set Mbps value.

Remember that the lower bit rate in the Vegas render settings is for Average bit rate, not minimum. Doesn't necessarily mean that it will render at that bit rate.

Remember that the lower bit rate in the Vegas render settings is for Average bit rate, not minimum. Doesn't necessarily mean that it will render at that bit rate.

My original question still stands. Averaging 28Mbps does not equal 11.5Mbps :) - Wouldn't you think that when setting the render to 28Mbps, the file should at least be close to that number? 11.5 is not even close. It's not even close to being half of the requested file size.

• Are you happy with the way the video from this encode looks?

If "yes", then we're done. Based on a recent thread, AMD VCE does a very good job with quality, speed and use of bits.

If "no", then select a different Encode Mode that will allow you set a constant bit rate that will make you feel more comfortable. Or select a different encoder that will allow you to change the group of pictures, number of P-frames, B-frames etc. The combinations are infinite, if you're willing to experiment.

• "Wouldn't you think that when setting the render to 28Mbps, the file should at least be close to that number?" 

          No.

Oh, it's your hardware encoder, which makes the unpredictability factor worse. See if checking two-pass gets you closer.

Oh, it's your hardware encoder, which makes the unpredictability factor worse. See if checking two-pass gets you closer.

I've used two-pass in the past - but - if you look at the first image I posted in my OP above - Two-pass is greyed out (can't be selected). Do you think that I might see better results with CPU only?

@Qoncussion I brought this issue up shortly after Nvenc hardware encoding was implemented. Probably similar for Vce.

To get the output bit rate you require just experiment by selecting a small loop region and keep modifying the settings. When you get close to your desired output save it as a new template.

Its as if hardware encoding is less linear and more like discrete steps sometimes.

The default settings in the hardware encoder render templates I think are more a guide than what you will actually get.

Non hardware encoding appears to be more linear, but it still won’t give you exactly what you enter.

@Qoncussion if your requirement is 28 Mb/s, use Magix AVC format, change the template to 'constant' bite rate and select 28,000,000. Render and done!

EDIT: Sorry I had too many things going on......I originally thought your requirement was needing 28 Mb/s, but reading through again it appears you just want to know why with the settings you selected you did not get 28 Mb/s, which others have already answered.

In spite of my low expectations mentioned earlier, when I ran a high motion video through the Magix AVC encoder in Vegas Pro 15 with the template set to match yours (AMD VCE 32/28 Default Peak Constrained), I got close to the target bit rate. Others could check Vegas 16.

I tested the red car bennchmark with the published templates, but I have no AMD.
Besides the default MC template I have for the supposed rendersettings two other possiblities, the Magix AVC QSV en NVENC options, which I used both.
Here are my screenshots of the rendersettings and the rendered MediaInfo ot the exportfiles
QSV

NVENC

I used VP16 and the same AMD VCE settings as John_dennis and Qoncussion with the the following bite rate settings (and of course different source media)......

Max 32,000,000 Avg 28,000,000, Overall bit rate = 22.6 Mb/s

Max 40,000,000 Avg 30,000,000, Overall bit rate = 25.2 Mb/s

Max 50,000,000 Avg 40,000,000, Overall bit rate = 29.7 Mb/s

I am wondering if there might be something funky with my AMD setup... I ran a test at ~ Max: 60,000,000, Avg: 50,000,000. It rendered at around 10Mbps (10,000,000). Not sure what's going on. I switched from AMD VCE to Mainconcept, and it rendered right in line with the settings

In VPro 16 the plain MC AVC cannot be used for GPU rendering, so all is done by the CPU, so also the results cannot be compared between the GPU- and only CPU renders.

I am wondering if there might be something funky with my AMD setup... I ran a test at ~ Max: 60,000,000, Avg: 50,000,000. It rendered at around 10Mbps (10,000,000). Not sure what's going on. I switched from AMD VCE to Mainconcept, and it rendered right in line with the settings

I read that your R9 290X has VCE 2.0. That's quite an early version and it's not really a surprise if there are quirks. VCE is up to version 4.0 now in Vega models. You might try a newer or older driver than your current one to see if that makes a difference.

Since the advent of on-cpu encoders (actually since cq/crf software encoding) average bitrates and thus file sizes have become really hard to nail down with any consistency.

Reason: There is no longer a need for a time-consuming scan pass, thus no chance for the encoder to anticipate and adjust its game plan. Having become a guess-and-go situation, it will probably get no better, except in deference to slower legacy software solutions, which retain more quality, as well.

It would probably be better if your render instructions said, "Up to 28Mbps," Also, the same advances in these newer encoders put the classic "more bits = more quality" cliche at risk of extinction. It ain't that way any more.

Posted without comment:

@john_dennis the best 2 readings you get for VCE, top of both columns, fall just above and below the reading I got for BruceUSA's sample rendered VCE clip. I guess given two different clip samples used, two different Radeon software versions it does show a fair bit of consistency in the VCE results.

BruceUSA

When I do any of these HORQM tests I always set the hardware, say Nvenc render settings at vbr best quality.  After all it is a comparison with CPU rendering although not exclusively, each other as well.

What's also important is how recent the VCE version is, as Nick mentions.  I'd imagine the BruceUSA version was most recent given he has a very modern rig.

I hope to get use of a buddys laptop next week that has Vce enabled, I think via the CPU.  If so i'll post and update my "General purpose" testing with that.  I'll post the VCE version in use too once I can establish it.

Update ... I couldn’t get Vce to render at a similar data rate to what was already done with the the other renders in “General purpose”

General purpose

Unfortunately while doing the BruceUSA tests I set the data rates for other renders a small bit higher than the VCE clip.  This needs to be taken into account.  Obviously I don't know what quality render settings were used for the BruceUSA VCE rendered clip.

Ive updated the BruceUSA info above with a laptop that's VCE version 3.1, the data rate is a bit higher.

Shows theres not that much difference between say Nvenc, Vce etc, its data rate dependent also.

Some updated tables ... https://www.dropbox.com/sh/xh8vkexgxif53m5/AAAH135JcFgIuqoQgFp5-o1za?dl=0

JN. The VCE rendered clip at default template. I will run some testing later on today between VCE 2.0 and VCE 4.0. I got two computers AMD system VCE 4.0 and Intel system with VCE 2.0 AMD R 9 290X.

@BruceUSA Truly excellent. I hope you realise what you’ve started here😀

@BruceUSA Truly excellent. I hope you realise what you’ve started here😀

Wow. I am speechless. :).

VCE 4.0 render settings @ Max 50,000,000. Avg 28,000,000. The file bit rate read by Mediainfo at 16.4 MB/s

VCE 4.0 render settings @ Max 28,000,000 Avg 20,000,000. The file bit rate read by Mediainfo at 15.1 MB/s

VCE 2.0 render settings @ Max 50,000,000 Avg 20,000,000. The file bit rate read by Mediainfo at 16.5 MB/s

VCE 2.0 render settings @ Max 28,000,000 Avg 20,000,000. The file bit rate read by Mediainfo at 16.3 MB/s

VCE 4.0 render Default settings Max 24,000/ Avg 12,000. The file bit rate read by Mediainfo at 10.6 MB/s

VCE 2.0 render default settings Max 24,000/Avg 12,000/ The file bit rate read by Mediainfo at 10.4 MB/s.

Seem to me the most you can get out of VCE is 16.4 MB/s with Avg 28,000,000/Max 50,000,000. Any higher will get error. I don't see the point of changing the default settings to any higher. I imported both files onto Vegas and set the compositing mode to different. The default settings video file look very good to me on both files VCE 2.0/4.0

The video file is from SONY Project Red Car.

@BruceUSA Indeed. The higher data rate (24,572) you produced with your benchmark was of course a 4K clip.