Vegas Pro... 8bit vs. 32bit Samples...
 

I just downloaded the Vegas Pro trial... I haven't seen a huge performance increase, but the it seems much more stable...

I did a little test to see the difference between the 8bit and 32bit.

I found I'd rather edit in 8bit and color correct in 8bit... rather than edit in 32bit... I get roughly 10-11fps playback at Full HD (and converting from 60i to 30p on the fly) vs. 1.5fps playback at 32bit... so 10% of the performance for slightly more accurate colors?

http://www.lousyhero.com/pics/dvi/8bit.jpg
http://www.lousyhero.com/pics/dvi/32bit.jpg
http://www.lousyhero.com/pics/dvi/8bitcc.jpg

First is 8bit
Second is 32bit
3rd is 8bit with a slight color correction filter dropped on (not tweaked)

Joe:

Conclusion ? 32 bit looks nice, but the color corrected version looks as good,too, IMHO.

What is difference in render times between the various scenarios ?

Looking at the luma histograms, the 32-bit ranges from 0 - 255 where as the 8-bit ranges from approx. 16 to approx. 250.

Is this an IRE setup thing + conforming (loosely) to the usual 16 - 235 range?

As a result, the 8-bit does look somewhat less vibrant.

The histogram for the color corrected version shows the black clipping strongly.

Of course, looking at an RGB bitmap can be misleading, anyway. The proof of the pudding is really in the YCrCb channels as far as validity of the pixel values is concerned.

Problem I have is while 32 bit looks better than 8 bit on the timeline it looks worse when rendered. Colors were blah and the black levels were wrong. Looks like I it would when one went from a 3 CCD SD cam to a single CCD SD cam.

Also when rendering 16x9 SD to mpeg 32 bit took 3 times longer and was not as good.

When rendering to wmv the quality was about the same but it took 4 times longer to render.

There may need to be a VP8-B or C to adjust how 32 bit renders?

Oh Yea, 32 bit makes previewing way too sluggish. I have a newly built fast system with quad core, 4 gig of ram and a good video card. I think it may be a while before I want to use the 32 bit thing...

Danny Fye
www.dannyfye.com
www.vidmus.com/scolvs

Can you disable 32-bit floating point (fp) processing in Vegas 8?

Whilst developing our software, I toyed with 32-bit fp vs. integer and found that it was slower for the same calculations. Primarily, this is because it takes a relatively long time for the CPU to convert to and from fp.

Yes you can disable 32-bit floating point processing in Vegas. Keep the project in 8-bit bit depth.

Some of the filters process in 32-bit float internally (in 8-bit projects).

2- Vegas has two different 32-bit modes:
A compositing gamma of 2.222. This behaves like old school Vegas.

A compositing gamma of 1.000. All the values are converted from gamma corrected values to linear light values for processing. This is slower. Filter output is different.

3- John, check your private messages... I'd like to bounce some computer programming stuff off you.

Thanks for the comparison Joe, would love to see the 32-bit CC to round it out.

Glenn,

That's critical information (1.0 vs 2.222 gamma).

If the gamma, gain and offset are properly set, the only differences between 8-bits and 32-float would be:

1) Less contouring and shallow gradients,
2) Less noise when between levels, and
3) Slower processing.

Oh well, you can't have it all...

That said, it might take a bit before we figure out that "properly set" part.

It may be better, but I think the point is, you can color correct 8bit to look better than 32bit and still have it preview at decent frame-rates.

All of my HDV stuff has that weird grey overlay over it... and I assume it's from lack of color-depth... I drop some color correction filters, and tweak them if I have time... I'm not overly concerned about accurate colors, I just want it to look better.

Heh, heh. The question in my mind is - will color corrected 32-bit / color corrected look better than 8-bit / color corrected. I'd do some testing myself, but I'm having issues getting V8 Pro to stay up and running on my system.

Sorry, I just found this thread.. So Vegas 8 just converts 8bit HDV footage to 32bit? I thought you would need to capture your clips to a higher bit rate in the first place using something like an intesity card and maybe the 10bit Cineform Neo HD codec, for 32bit to actually make any difference?

I do not have my copy of Vegas Pro 8 yet, so I have not done any testing.

Floating Point speeds on different hardware (different computer CPUs, different brands (AMD, Intel, others) is dramatically different.

So, it will be appropriate to post specific computer details with each post referring to performance timings.

I mistakenly stumbled into this last night after setting my properties to 32bit instead of 8. Took 6hrs 3mins to render a 55min sequence of HD on a Quad Core with 4Gb of RAM. Once I sussed out what I had done, I had a look at the difference and was pretty surprised.

The "grey cast" that Joe mentions in his second post is plain to see in the Comparison 2 grab.

Did some digging and found this thread.

Decided to try some color correcting (Levels with some tiny tweeks to Input start and Gamma) to try and match the 8bit to the 32bit and got "close but no cigar". Seems to have a little less gloss and the whites don't "pop" as much.

Would be interested to hear what the "Big Guns" would suggest to get a better match.

Just done a side by side of Joes pics as well for ease of viewing.

I think you guys are a little confused. Vegas doesn't convert 8-bit data to 32-bit. Universal law-you can't get something from nothing. What Vegas does do is process(as in calculate) the data with 32 bit floating point math precision , instead of 8bit fixed point math. Yes, you get more accuracy(as in decimal places) and less banding.

also, the color bandwidth you see out of a render is dependent on a whole lot of things besides 8-bit or 32 bit processing. Are your monitors calibrated? What codec did you render and decode with? All of these things will affect the black(shadow) and white(hi-lite) values you are looking at. 32-bit seems to affect that because it changes the way a codec handles the Y'PrPb conversion to RGB and back. Each codec is different. You can't decide the effect you see, haziness or crushed shadows, is the fault of 32-bit processing alone.

Just to make an important distinction: the calculations themselves are not 8-bit. They are almost certainly either 16- or 32-bit integer - the latter being most likely. This is what processor extensions such as SSE/SSE2 etc use. Performing true 8-bit calculations is a very inefficient and slow method (ironically). In essence, the incoming 8-bit frame is converted to 32-bit (integer). Once the calculations are complete for a given effect, the new frame is then converted to 8-bit.

For single effects/transitions etc, the 8-bit mode perfectly adequate. For multiple effects on the same frame, that's where the 32-bit floating point becomes a benefit. All the intermediate frames are kept at 32-bit floating point. The downside to 32-bit floating point is that it is very CPU intensive. Beyond simple add/subtract/multiply, calculations can't be done using SSE/SSE2/SSE3/SSSE3 etc. They have to fall back to the ancient x87 co-processor which is v-e-r-y slow.

John..Thanx. My understanding is, obviously, incomplete. Thanx for speaking up...at last, someone who knows what they're talking about.

As one of the ones that "doesn't" know what he is talking about, all I can say is to my humble eyes, the 32bit looks a better image. This then spurred me on to correct my 8bit one to look more like it.

Alastair...

I'm sorry...didn't mean to be insulting. There's so much misinformation coming from people who think they know, myself included in this category, that it's nice to get an explanation that is technically correct.

In the end, the best judge is one's own eyes. Or the customer's ;o)

The dramatic difference is because some (but not all) codecs change behaviour depending on whether the project is 8-bit or 32-bit. This is not that intuitive in my opinion, but is what Vegas does.

If you apply a "studio RGB to computer RGB" color corrector preset onto a clip, it should give the same effect. This is assuming your clip is a codec whose behaviour changes, e.g. HDV or MPEG-2. Also, be in 2.222 compositing gamma (not the other one).

2- The problem originates because of how Vegas is designed... in many cases, Vegas doesn't show the image as it should look in the preview. The Video Preview window is often wrong. And this problem was there all along, before Vegas 8.

3- Some information on how to get your levels correct:
http://glennchan.info/articles/vegas...or/v8color.htm

And now you just compound it further by spelling me name wrong!

None taken. I am most definetly in the "throwing rocks at the moon" camp when it comes to getting a handle on this. Like I said, my mistake just led me somewhere interesting, that may end up helping me get better looking footage. Plus I did ask for the "Big Guns", so I guess I got what I asked for.

So....I'm happy!

Seems like the gun always goes off just when it's pointed at my foot!
Not so sure how big my gun is...LOL.
Anyway, it's been corrected.

Glenn are you serious??
 

2- The problem originates because of how Vegas is designed... in many cases, Vegas doesn't show the image as it should look in the preview. The Video Preview window is often wrong. And this problem was there all along, before Vegas 8.


How then is it possible then to colour correct in Vegas?? If you can't trust what Vegas spits out?

If it's SD, check it on a properly calibrated monitor.
If it's HDV (different colour space) your guess is as good as mine. Most folks are using their HD TV as a reference.

I tend to rely, pretty heavily, on the WFM to set black and hi-light levels. This methods works quite well for me. Colors are a little more difficult, but, again, applying the right FX level helps a lot. The only thing the Preview window is really misleading about is the sharpness of the image. I tend to oversharpen footage, because Vegas doesn't show previews at the real resolution.

You need to manually set things up correctly.

For DV footage, preview via firewire out to a broadcast monitor (you have to calibrate the broadcast monitor). Use the default Vegas codec.
You can also use the Windows secondary monitor as a preview... check color management, and check the studio RGB box (in the same preferences page).

For HDV footage, either set the Windows secondary preview up correctly (you have to refer to the table). Or, go HD-SDI out.

2- You could also apply a "studio RGB to computer RGB" preset to the video preview level to get things right in certain situations (e.g. when working with DV footage) and looking at the video preview window... but this is dangerous if you forget to take it out.

?
You can set the video preview window to show full resolution (don't use the draft mode or half-resolution modes), make sure that the video isn't stretched to fill the video preview window, use the (full) preview modes, and turn off simulate device aspect ratio.

Granted, in some cases you can't have it all (fast preview, correct PAR) and get resolution looking as it should be (1:1 pixel mapping).

Thanks Glenn
 

as usual always prompt and willing and helpful, I know your efforts are greatly appreciated on this forum.

Regards

Adam

And as well Bill and Mike
 

you've been great as well, just some of the many people that make this forum so worth while!

Cheers

Adam

Glenn...

Even with these settings, the images are blurry and out of focus when viewed via a Windows Secondary Display. The same footage is razor sharp when viewed on the same monitor on an application outside of Vegas.

Adam...
Thanx, mate.

So, (please excuse the cameraman who's not an engineer)! would V8 let you edit 8 bit footage in 8 bit and any transitions, effects and CC occur in 32bit? Or does it have to be all one and not part of the other? From a discussion currently ongoing in the JVC GYHD forum Paolo Ciccone recommends editing in 8 bit then export to AE for example and carry out any compositing and CC work in 32bit. Reason being any CC work done to 8bit in 8bit tends to impair the original work.

I've been looking at maintaining the image quality right to the end but have been finding any CC work or compositing does slightly impair the result. I'm now considering using something like Adobe AE (if this is compatible with VEgas???) to carry out anything additional to normal cuts. However, having not yet installed the V8 upgrade from 7 can I save myself the cash and use V8 to do all CC work in 32bit?

Stuart

Stuart...
Vegas Pro v8b will process in 8-bit or 32-bit, your choice. Within 32-bit processing, you also have the choice of gamma 2.2 or a linear gamma to improve transition fades, among other things.

Unfortunately, using 32-bit is a wee bit tricky. One needs to carefully adjust FX settings (primarily the Levels FX) to realize the highlight recovery. Glenn Chan has a few very good articles explaining the process,but, the editor still needs to wrangle Vegas to use it to its best potential.

Read John Miller's post in this thread. It's very informative re: what 32-bit floating point processing means.

Thanks Bill,
seems I need to learn more on this before I use V8 to do anything in 32bit. I'll read the articles!
Thanks for the advice though! maybe using After Effects would be a simpler option for 'non engineer types' like me!
Stuart

Ack, my mistake! I was thinking about the Video Preview.

For square PAR formats: You could uncheck the box which says "scale output to fit display" in the preferences for Preview Device. This doesn't work for DV though.

Affect Effects has a few of the same issues and a few different issues (Quicktime color management gone awry, not being able to see superwhites).