Page 1 of 1

varying host performance over sections of a clip??

Posted: Mon Oct 31, 2016 10:18 pm
by Zach
I've been trying to get some performance metrics from my machine, in a timely manner, but it seems that idea (the timely part) is out the window. Has anyone noticed a vast performance difference between different sections of a clip?

What I mean is I'm seeing something weird. I'm basing my tests off the first episode of an old hand drawn anime (Lodoss War). I thought I could shorten the test by just running through the opening credits, but this appears to not be the case.

Test Conditions:
Tests conducted with latest Neat Video DEMO version for all hosts.
Same Noise / Filter profiles loaded for all hosts
Same clip section used.
Source was encoded as 1436x1080 HuffYUV YUY2 / 4:2:2
No other filtering or processing effects were used for the tests.
All render tests were saved out as UT Codec YUV4:2:0 BT.709 / 1436 x 1080
No audio processed.
GPU Only Processing using an HD7870 & HD7950 together.

Test Hardware:
i5 Ivybridge turbo'd up to 4.1Ghz
16GB RAM
HD 7950 & HD 7870 installed.
Read Drive: 100MB+ R/W speed.
Write Drive: RAID 0 array capable of 100MB+ R/W speed.
Windows 10 x64

All hosts I ran the test on pretty much had the same render time for the opening credits, yet if I were to run the whole episode through Neat Video, suddenly there are much larger gaps between hosts.

both Virtualdub x32/x64 and Virtualdub FilterMOD x32/64 render within reason of each other. Vdub seems to be slightly faster. The entire 24 minute episode took about 2h:06m to filter and save as UT YUV420

MAGIX Movie Studio 13 Platinum took 1h:28m

Adobe Premiere CC 2015.4 took roughly 2h:28m

Yet when I ran just the opening credits through NV, all hosts had virtually identical render times within ~45 seconds of each other.

I'm wondering if anyone can explain why it comes out that way? Is it down to differences in the way NV is coded per host? The performance of the host programs plugin system?? If it was the host itself, I would expect to see equal speed issues on a short vs long clip, no ?

This is more curiosity than anything. But it also factors into my decision about what host application I want to invest in next (moving away from Vdub), and likewise, purchase a new Neat Video license for.

Even though I have "issues" with some of the aspects of Movie Studio 13, it does have a nice interface (although severely cut down from the Vegas PRO branch) so I am kind of leaning that way. It's pretty affordable (although I wonder if a v14 is on the way) and seems to give me the best performance (although again I am hoping to upgrade hardware in the future).

Posted: Tue Nov 01, 2016 12:48 pm
by NVTeam
Generally speaking, there are always differences in performance just because of different host applications. Different host applications use different internal processing with different optimization. Even if Neat Video itself runs with the same speed (it usually does, if the filter settings and input data are the same), the overall render speed may vary because of the host-side differences. If you take the speed of Neat Video itself (as measured by its Optimize Settings tool) as 100%, then the speed of the whole render may vary (depending on host application, as well as input and output codecs) from say 25% to 75%. If you use a complex codec to decode and encode the video, then the speed loss is high. When you use a simple codec, the loss is lower.

I am not sure why a part of the clip would be processed without much speed difference but the whole clip would show more difference. I can guess that perhaps the output codec is doing more work (so taking more time) when processing frames with richer content and more action. The credits may be simply too easy (and fast) to process, hence no differences in speed.

Generally I recommend to select the editing application that you are most comfortable with. The speed is of course important but it may change depending on a codec and other things (like other effects you apply), so I would not decide based on speed alone. The speed may improve when the codec is updated for example.

Hope this helps,
Vlad