Log in

View Full Version : DGNV Performance Results

JoeH
19th January 2010, 17:45
This is a thread to post performance results for DGNV tools. Examples could be performance of DGNV vs. DirectShowSource, NVidia GPU comparisons, etc.

@neuron2 - if you have any comments from your experience or simply from a "theoretical" level we would love it if you would share them.

Here are some initial results from my tests. Hardware: Core i7 920 overclocked to 3.8GHz, NVidia 9500 GT GPU.

In both cases I am encoding from Sony Vegas to X264. To do so I use DebugMode to frame serve to MeGUI 3.2. I use the 2 pass Bluray profile, with 20,000 bitrate and the speed set to "Faster". I will be comparing the use of an the UTVideo lossless codec to frameserving the original H264 file into Vegas using DGNV.

UTVideo RGB - Intermediate codec (lossless conversion of the original H264 file)
1st pass 30 fps
2nd pass 20 fps

DGNV frame serving the original H264 file into Vegas using AVFS
1st pass 36 fps
2nd pass 26 fps

Conclusion: Frame serving H264 into Sony Vegas using DGNV is between 20% and 25% faster than using UTVideo RGB lossless codec. I should note, however, that Vegas paints the thumbnails slower with the DGNV solution (possibly because of the use of AVFS) than with UTVideo codec.
The Scientist
19th January 2010, 18:14
[B]... Core i7 920 overclocked to 3.8GHz, NVidia 9500 GT GPU.Wow..... don't wanna swap your system for my wife do ya? :)
poisondeathray
19th January 2010, 22:28
A bit off topic, but if you export a lossless intermediate out of vegas, I suspect it will go faster because vegas only has to render effects/edits once, instead of twice when using debugmode (since you are using 2passes in megui)
JoeH
22nd January 2010, 17:45
A bit off topic, but if you export a lossless intermediate out of vegas, I suspect it will go faster because vegas only has to render effects/edits once, instead of twice when using debugmode (since you are using 2passes in megui)

It is faster, but then I would need to recode it into H264 anyway using MeGUI, which would make the process as a whole longer. Honestly, DebugMode is incredibly light weight and makes very little difference. For example, it is much, much faster to encode to H264 from Vegas using MeGUI through DebugMode than using the SonyAVC codec included with Vegas (and the quality is higher with MeGUI as well).

That having been said, there is some overhead. If I take an H264 file and encode it straight into MeGUI it will be 20% faster for the 1st pass and 10% faster for the 2nd pass than if I run it Vegas - DebugMode - AviSynth- MeGUI. Not much difference all considered.

I hope to post some comparisons soon of DirectShowSource vs DG-NV, both straight into MeGUI without passing through Vegas.
poisondeathray
22nd January 2010, 18:34
It is faster, but then I would need to recode it into H264 anyway using MeGUI, which would make the process as a whole longer. Honestly, DebugMode is incredibly light weight and makes very little difference. For example, it is much, much faster to encode to H264 from Vegas using MeGUI through DebugMode than using the SonyAVC codec included with Vegas (and the quality is higher with MeGUI as well).

That having been said, there is some overhead. If I take an H264 file and encode it straight into MeGUI it will be 20% faster for the 1st pass and 10% faster for the 2nd pass than if I run it Vegas - DebugMode - AviSynth- MeGUI. Not much difference all considered.

I hope to post some comparisons soon of DirectShowSource vs DG-NV, both straight into MeGUI without passing through Vegas.


I don't think I explained it very well

When you do more than simple cuts editing in vegas (i.e. you use filters, color correction etc...), it is very very slow to render. Instead of doing this twice (when you use debugmode to frameserve to megui), you do it once. You will see it is much much faster encoding to a lossless intermediate. It's the same idea with heavy avisynth filters. The only time this doesn't make a difference is if you do a 1pass encode like crf.

Anyways, thanks for the data and observations

Cheers
JoeH
29th January 2010, 17:59
Thanks for the advice, poisondeathray. I don't use many filters right now in Vegas, but if I do in the future I'll keep it in mind.

Here are some more performance data: With my 9500 GT, I get about 40fps maximum playback speed within DGIndexNV when I play 1080p H264 videos.

What type of numbers do others get with more or less powerful GPUs?