Vegas 15 GPU Acceleration Testing

[EEVblog]

Tests for quality differences:
Using the NVENC engine and not
https://en.wikipedia.org/wiki/Nvidia_NVENC
All in 1080p50 12Mbit/28Mbit

Results:
NVencoder HQ : 1min 47sec 179MB
NVencoder LQ Default : 1min 43sec 180MB
MainConcept : 2min 41sec 192MB

[cdev]

On the GPU acceleration video, the audio sounds horrible through my Sony MDR-v6 headphones.

[EEVblog]

On the GPU acceleration video, the audio sounds horrible through my Sony MDR-v6 headphones.

I can't hear any difference on my reference monitors or headphones.

[cdev]

It seems to me as if its just the beginning part of first video "Vegas 15 GPU Acceleration Testing" that has a kind of raspiness in my headphones. The later part sounds okay, not raspy but it does sound as if you equalized it using a U shaped curve - it sounds "thin" in the middle, it maybe needs a bit more "presence" it might sound better with more of the middle frequencies, the ear is the most sensitive to sounds around 1 khz?

The others sound okay.

If you had a particular audio clip that you could use for testing at a set level, then you could analyze its spectral content to see if its spectral content looks out of the ordinary in any way.

It may well just be me.

These programs are all really cool in their own ways, and have the capability to analyze the spectral content of audio..

http://www.sonicvisualiser.org/ which is extremely sophisticated and has all sorts of plugins available for it

Baudline. http://www.baudline.com/index.html

and Speclab http://www.qsl.net/dl4yhf which you can also use as a VLF receiver.

[EEVblog]

So I installed a GTX-1050 card with the latest NVIDIA chipset and NVENC encoder and I got zero difference in the render time compared to the GTX750Ti using Vegas 5 that supports the NVENC encoder.
Very surprising!

[cdev]

Strange.  Is there any chance that its not using either GPU at all!???

[dr.diesel]

I suspect it's not utilizing the GPU at all.  I don't know how to do it on Windows, but I'm sure there is a utility to view GPU processes/usage in real time, and an idle GPU card here would explain it.

[thm_w]

Thats quite surprising, 1050 Ti peak power consumption might be a few watts lower than 750Ti but its not much.
Although, you can use the 1050 Ti to mine ethereum at ~13MH/s, 750 is less than half of that.

I suspect it's not utilizing the GPU at all.  I don't know how to do it on Windows, but I'm sure there is a utility to view GPU processes/usage in real time, and an idle GPU card here would explain it.

Watch the previous video at the top of this thread, he monitored GPU usage and it was at 100%. When using cpu it was below 20%. So I'd be very surprised if he put in a new card and it was any different.

edit: didn't realize its a 1050 not a 1050 Ti, which is a terrible card (you could see the passmark was almost the same as 750ti). Although not sure if it matters for GPU rendering.

[SparkyFX]

GPU accelerated video rendering is a hot mess since years. As far as i understood it NVEnc is a nVidia only thing will only utilize GPU, but not the CPU. Handbrake H.264, FFmpeg, and according to Wikipedia allegedly MAGIX Vegas too, support OpenCL and could theoretically utilize CPU and GPU (even on nVidia GPUs with latest drivers), but no idea if the encoder can or does select that.

The thing is, the libx264 (used in handbrake and ffmpeg,...) does some lookaheads via OpenCL, but not the main encoding cause they did not want to rewrite the whole code. So it won´t drastically benefit from it. Handbrake can move scaling to OpenCL, ffmpeg can filter with it, but itself can not beat libx264 to fully use OpenCL. So "support for OpenCL" is to be taken with caution, although the list of encoders with OpenCL support is long.

Handbrake and ffmpeg however also support NVEnc. Why mention ffmpeg? Because it always represents a good state of open source support for these things and is actively developed.

AMD allegedly worked on a fork/version of h.264 that uses OpenCL, but nothing is to be found except some interviews and an old anandtech article.