Results of render times for ALL Vegas fX

johnmeyer wrote on 7/9/2005, 5:46 PM
Over the past few years, a lot of people ask why it takes so long to render. In many cases, it turns out they are using various fX, some of which take more time than others to render. Of course, there are various bus fX (supersampling and motion blur) and also compositing, all of which can consume even more CPU cycles. However, I still thought it might be useful for the Vegas community to have a benchmark, based on measurements, of which each fX takes to render.

I put forty-eight identical ten-second NTSC DV AVI clips on the timeline. I then assigned to each, one of the built-in Sony fX. I used Vegas 5.0d, although I would expect the relative results would be similar in Vegas 6.0b.

I then created a script that assigned a region to each event and named that region with the fX name. I then used the batch render GUI script to render each ten-second clip to its own file. I then copied the file listing of all these files to a spreadsheet and used the difference in the creation timestamps to compute the duration of the render for each clip. I sorted that spreadsheet in two different ways, and present the results below. The actual times, in seconds, is not that useful since your computer will be different than mine. However, the relative times are probably the same, no matter what computer you have. As you can see, there is a HUGE difference in rendering times depending on the fX.

Using this information, you can make informed decisions on trading off creativity against whatever deadlines you have to get the job done.

I could upload the Excel spreadsheet to one of the Vegas support sites, if anyone is interested.

Below, I duplicated the same information twice: once sorted alphabetically by plugin name, and then sorted by performance starting with the fastest and then going to the slowest.
Plugin fX			Time (seconds)	"x" times (relative to fastest)
Sony (Legacy) Broadcast Colors 13.9 1.1
Sony Add Noise 64.8 5.1
Sony Black and White 16.2 1.3
Sony Black Restore 12.7 1.0
Sony Border 19.7 1.5
Sony Brightness and Contrast 15.0 1.2
Sony Broadcast Colors 18.5 1.5
Sony Bump Map 99.5 7.8
Sony Channel Blend 18.5 1.5
Sony Chroma Blur 50.9 4.0
Sony Chroma Keyer 59.0 4.6
Sony Color Balance 19.7 1.5
Sony Color Corrector 22.0 1.7
Sony Color Corrector (Secondary) 23.1 1.8
Sony Color Curves 19.7 1.5
Sony Convolution Kernel 79.9 6.3
Sony Cookie Cutter 19.7 1.5
Sony Deform 302.1 23.7
Sony Film Effects 358.8 28.2
Sony Film Grain 42.8 3.4
Sony Gaussian Blur 88.0 6.9
Sony Glow 110.0 8.6
Sony Gradient Map 26.6 2.1
Sony HSL Adjust 42.8 3.4
Sony Invert 17.4 1.4
Sony Lens Flare 75.2 5.9
Sony Levels 13.9 1.1
Sony Light Rays 229.2 18.0
Sony Linear Blur 163.2 12.8
Sony Mask Generator 12.7 1.0
Sony Median 381.9 30.0
Sony Min and Max 2,950.2 231.7
Sony Mirror 49.8 3.9
Sony News Print 30.1 2.4
Sony Pinch Punch 86.8 6.8
Sony Pixelate 22.0 1.7
Sony Quick Blur 41.7 3.3
Sony Radial Blur 79.9 6.3
Sony Saturation Adjust 44.0 3.5
Sony Sepia 17.4 1.4
Sony Sharpen 50.9 4.0
Sony Spherize 82.2 6.5
Sony Swirl 98.4 7.7
Sony Threshold 18.5 1.5
Sony Timecode 16.2 1.3
Sony TV Simulator 69.4 5.5
Sony Unsharp Mask 144.7 11.4
Sony Wave 71.8 5.6


Here is the same table, sorted by speed:

Plugin fX			Time (seconds)	"x" times (relative to fastest)
Sony Black Restore 12.7 1.0
Sony Mask Generator 12.7 1.0
Sony (Legacy) Broadcast Colors 13.9 1.1
Sony Levels 13.9 1.1
Sony Brightness and Contrast 15.0 1.2
Sony Black and White 16.2 1.3
Sony Timecode 16.2 1.3
Sony Sepia 17.4 1.4
Sony Invert 17.4 1.4
Sony Broadcast Colors 18.5 1.5
Sony Channel Blend 18.5 1.5
Sony Threshold 18.5 1.5
Sony Color Balance 19.7 1.5
Sony Color Curves 19.7 1.5
Sony Cookie Cutter 19.7 1.5
Sony Border 19.7 1.5
Sony Color Corrector 22.0 1.7
Sony Pixelate 22.0 1.7
Sony Color Corrector (Secondary) 23.1 1.8
Sony Gradient Map 26.6 2.1
Sony News Print 30.1 2.4
Sony Quick Blur 41.7 3.3
Sony Film Grain 42.8 3.4
Sony HSL Adjust 42.8 3.4
Sony Saturation Adjust 44.0 3.5
Sony Mirror 49.8 3.9
Sony Chroma Blur 50.9 4.0
Sony Sharpen 50.9 4.0
Sony Chroma Keyer 59.0 4.6
Sony Add Noise 64.8 5.1
Sony TV Simulator 69.4 5.5
Sony Wave 71.8 5.6
Sony Lens Flare 75.2 5.9
Sony Radial Blur 79.9 6.3
Sony Convolution Kernel 79.9 6.3
Sony Spherize 82.2 6.5
Sony Pinch Punch 86.8 6.8
Sony Gaussian Blur 88.0 6.9
Sony Swirl 98.4 7.7
Sony Bump Map 99.5 7.8
Sony Glow 110.0 8.6
Sony Unsharp Mask 144.7 11.4
Sony Linear Blur 163.2 12.8
Sony Light Rays 229.2 18.0
Sony Deform 302.1 23.7
Sony Film Effects 358.8 28.2
Sony Median 381.9 30.0
Sony Min and Max 2,950.2 231.7

Comments

Spot|DSE wrote on 7/9/2005, 6:00 PM
Excellent info, John! Thanks for sharing. A couple of those surprise the heck outta me! I've never tested all of these, but would have assumed (wrongfully) that some of the grains would be at the bottom of the speed list too. I also would have thought Median to be the worst of all filters in terms of speed, but was wrong about that, too.
Excellent stuff!
johnmeyer wrote on 7/9/2005, 6:26 PM
Spot,

The one thing I didn't mention is that it is possible that the performance may depend somewhat on what values I set in each filter. I don't think this matters in most cases, but it might. I just chose what I felt was a "representative" preset for each fX.

Also, I just finished updating the formatting. If you looked at it a few minutes ago, the columns still weren't straight. They are now.

JohnnyRoy wrote on 7/9/2005, 8:34 PM
Yea, really excellent work John! I thought Median would be the biggest hog too but Min Max blows it away by a long shot. This is good stuff. Thanks for sharing it with us.

~jr
Grazie wrote on 7/9/2005, 11:58 PM


JM! Excellent piece of work! IF this hasn't been posted before, in such a clear and transparent way, then it should have been.

We now have a table that we can all "link" to when a Newbie OR Oldie comes to this point in their development.

WELL DONE JOHN!

What we now need is a tiny readout, on an effect menu, that will inform us how long this will take! HAH! Nah, Sony, only jossin' ! Well kinda . . . .

I'm now wondering will the order in which I/we "chain" FXs make a render time change? Blur before CC? CC before Blur? The reason I ask is that I can "change" the look of my piece if I swap the FXs around - yeah? Soooo... this would imply, to my non-programming mind, that there COULD be a total render-time difference - yeah?

Only a thought ..

Grazie

DJPadre wrote on 7/10/2005, 1:50 AM
wow, im sure i speak for everyone when i say that this is much appreciated.

This is something i would have though Sony would have released as a white paper, but even so, this information has been greatly needed
MarkWWW wrote on 7/10/2005, 2:49 AM
Excellent work John, many thanks for doing it. I've had a vague intention to do something like this for a while now but I've never got round to it - now I don't have to. :-)

As you say, the actual times will vary from machine to machine depending on the CPU used but the relative times should be pretty constant across the range of machine speeds. But it might be interesting to see if there is any variation in the relative times between different generations of processors (e.g. between Pentium the III, Pentium 4, and Pentium-M series), and also between Intel- and AMD-sourced CPUs.

Also, I think it would be interesting to know how the times for each FX compared to the time for a region to be rendered without any FX processing. It seems to me that this "no-FX" time would be a more suitable choice for the baseline figure to calculate the "x times (relative to fastest)" column from.

Now we know not to use the "Min and Max" FX unless we really, really, need it. Actually I don't even know what "Min and Max" does - I'm off to have a look at it now. It had better be really good to justify all that time. :-)

Anyway, thanks again for doing this work.

Mark
Joe Lombardo wrote on 9/1/2005, 7:39 AM
This is an incredible resource. Thank you for all the time & effore it must have taken to produce this.
Dan Sherman wrote on 5/4/2006, 4:51 AM
Amazing information.
It is "nailed to the wall" for future reference.
Thanks, John, for your generosity in doing this testing and documenting the results.
Many of us will benefit,--- making informed choices when it comes to choosing FX.

craftech wrote on 5/4/2006, 5:08 AM
WOW!

That was a lot of work. Thanks you so much for taking the time to do that for everyone. Your contributions to this forum and fellow Vegas users is invaluable.

Regards,
John
MRe wrote on 5/4/2006, 5:44 AM
Excellent job johnmayer! Would it be possible to have the .veg available from some site?

This is also something I would like to have Sony's comments on (i.e. what is the reason for this, isi just because something requires more processing power or is it just because some plugins are poorly written).

Also reference data from V6 would be appreciated, eventhough I do not believe to see much difference there because the plugins are most probably the same, or?
ForumAdmin wrote on 5/4/2006, 8:53 AM
"it is possible that the performance may depend somewhat on what values I set in each filter. "

This is most definitely the case-more processing routines = more time. Try low and high settings of anything involving blur as an example.
johnmeyer wrote on 5/4/2006, 11:51 AM
Would it be possible to have the .veg available from some site?

I can't find it. If I do, I'll post it.

"it is possible that the performance may depend somewhat on what values I set in each filter.

I mentioned that in my second post above.
DGates wrote on 5/4/2006, 12:22 PM
That's good info.

I love the "Glow" effect in a lot of my work. But if I'm in a hurry, and I need to get a project out the door, I will sometimes forego it because I don't have the time to sit around while it renders.
johnmeyer wrote on 5/4/2006, 2:48 PM
Since I couldn't find the VEG file that I used to create the test, I simply created a new test, this time using Vegas 6.0d. I uploaded the VEG along with the Excel spreadsheet showing the results on my computer. This link is good for seven days or twenty-five downloads, whichever occurs first. If someone wants to host these files more permanently, be my guest. Here's the link:

fX VEG and XLS file (in Zip format)

Here is the data that is contained the Excel file. There are no big surprises compared to the earlier test, although a few fX did move to different positions in the list. I think this has to do almost entirely with the specific settings I used with each fX (especially the news print, which is quite different), rather than with any changes in Vegas itself since 5.0d.
                                 Render TimeRelative
Plug-in (seconds) Time
Black Restore 10 1.0
Brightness and Contrast 10 1.0
Mask Generator 10 1.0
Threshold 10 1.0
(Legacy Plug-In) Broadcast Colors 11 1.1
Levels 11 1.1
Sepia 12 1.2
Black and White 13 1.3
Timecode 13 1.3
Color Corrector (Secondary) 14 1.4
Invert 14 1.4
Border 15 1.5
Channel Blend 15 1.5
Color Balance 15 1.5
Broadcast Colors 16 1.6
Color Curves 16 1.6
Cookie Cutter 16 1.6
Color Corrector 18 1.8
Gradient Map 19 1.9
Pixelate 24 2.4
Film Effects 26 2.6
HSL Adjust 31 3.1
Quick Blur 32 3.2
Saturation Adjust 37 3.7
Mirror 38 3.8
Sharpen 40 4.0
TV Simulator 41 4.1
Chroma Blur 42 4.2
Film Grain 44 4.4
Chroma Keyer 52 5.2
Add Noise 53 5.3
Wave 58 5.8
Lens Flare 62 6.2
Radial Blur 63 6.3
Spherize 66 6.6
Convolution Kernel 68 6.8
Pinch-Punch 69 6.9
Gaussian Blur 77 7.7
Bump Map 80 8.0
Linear Blur 81 8.1
Swirl 83 8.3
Glow 92 9.2
Unsharp Mask 127 12.7
Light Rays 201 20.1
Deform 255 25.5
Median 728 72.8
News Print 1336 133.6
Min and Max 1723 172.3


* Performed with Vegas 6.0d, RAM Preview 156 MB
Video Preview On
2.8 GHz Pentium P4, Single Thread


Chienworks wrote on 5/4/2006, 4:59 PM
Thanks John!

If you don't mind, i took the liberty to put your file here:
http://www.vegasusers.com/testbench/files/john_meyer-vegas_6.0d_fx_render_results.zip
johnmeyer wrote on 5/4/2006, 5:30 PM
Great Kelly. Thanks for doing that!
Jay-Hancock wrote on 6/28/2006, 9:49 PM
I found another significant variable about FXs that greatly influences render times - when using a multiprocessor or multi-core system.

When you open the dialog for choosing a track or event FX you'll see the little icons have a green portion in the upper left, and a few (mostly third party FXs) have yellow instead of green. For the Sony FX's, only "film effects" has yellow. But nearly all the 3rd party ones that I know of have yellow in the icon.

Green means "multithreading friendly." Yellow means "single threaded only." If you add a single-threaded FX to the chain, it means that during the portion of your timeline that the single-threaded FX is active, only one core or processor will be used for the render and This isn't necessarily a big impact. If you use a SpiceMaster transition (for example), you only lose the use of your multithreading speed during the short duration of your transition (no big deal). If, however, you used a single-threaded, third-party deinterlacer FX on your video, the entire render will only use one core or processor. (I have only tested this with one single-threaded FX but I assume it holds true for all of them).

I learned this when I went to render a 2-minute section of an HDV project with the single-threaded smart smoother FX on a video track. The first few seconds used both cores, but then when it started writing to the hard drive one of the cores dropped out and remained idle trom then on. After reaching the 3% mark on the rendering dialog, the dialog predicted it would take 6 hours to finish. And the prediction was still climbing upwards (not down). I cancelled the render and switched off the single-threaded FX. Now both cores are continuously pegged at 100% and the render is going to finish in 2 hours.

Definitely something to consider if you are using third party plugins on a multi-core or multi processor system. (Most all the built-in Sony FXs are multithreaded).
Grazie wrote on 6/28/2006, 10:42 PM
"I cancelled the render and switched off the single-threaded FX. Now both cores are continuously pegged at 100% and the render is going to finish in 2 hours."

Great bit of detective work there, Jayster. Is this written down somewhere?

Thanks for that,

Grazie

Jay-Hancock wrote on 6/29/2006, 7:51 AM
Grazie - thanks for the acknowledgement. And really big kudos to John Meyer for this entire posting.

I had remembered that in the Vegas 6 manual (p. 338, under "Video Tab", or just search the pdf for "thread") it says the icon means an FX isn't multithreading capable. I didn't think about it much until I observed the actual impact of it. (I had figured it would only impact that FX, but I was wrong. It impacts everything!).