Render test (NTSC vs PAL)

eheh wrote on 11/30/2002, 2:50 PM
I got this new P4 2.53G, 512M PC2700, Asus PE4E, 80G WD SE as boot disk and 120G WD SE as video disk, Radeon 9000 Pro. Lovely system.

I did the rendertest using PAL templates (proerties and render) and was very worried when the render test did 3:55, no matter the configuration.

I then tried just to change preferences to NTSC and got the proper 2:12.

What's going on?
Is Vegas optimized for the NTSC market?

Comments

Spot|DSE wrote on 11/30/2002, 2:59 PM
I'll bet you changed the render settings from "BEST and RESAMPLE ALL" to just a normal GOOD/no resample, no blur....That would account for a HUGE difference. just did the same render on my machine in PAL, and it was only 2 seconds different than the NTSC render. But that was with all settings being exactly the same, except for PAL vs NTSC. Best/Resample/Blur all set to the same setting. Look deep at what's happening in that file, you'll see there are a LOT of proc-hitting settings in there. I wanted to really nail any proc with this test.
eheh wrote on 11/30/2002, 3:27 PM
SPOT, I looked deep - No differences that I can see.
Can I upload you the PAL and NTSC settings for you to check.
Where to?
Thanks
Wannabe wrote on 11/30/2002, 3:38 PM
I mailed em' to DSE@sundancemediagroup.com. Thats the address on his website
Spot|DSE wrote on 11/30/2002, 5:23 PM
I'm getting 4:02 on an AMD 1.4 when rendering to NTSC, and 4:19 when rendering your emailed veg to PAL. So it's slightly faster in NTSC, but not enough for any concern. I'm sure some of the SOFO folks can answer why.
There is greater vertical resolution in PAL, I'm assuming that's why it's slightly faster in NTSC. I could be wrong.
eheh wrote on 11/30/2002, 5:34 PM
Data amount wise they are nearly identical
PAL is 720[horizontal]x576[vertical]x32[bit/pixel]x25[frame/sec]
NTSC is 720[horizontal]x480[vertical]x32[bit/pixel]x29.97[frame/sec]
They calculate to 331,776,000 (PAL) and 331,444,224 (NTSC)
Quite the same.

I sent the files to DSE
Spot|DSE wrote on 11/30/2002, 6:07 PM
Spot=DSE, for what it's worth. My nickname is Spot because of my last name.
Anyway, I don't know why it's taking so much longer for you to render the PAL vs NTSC. Maybe Dither or one of the other SOFO guys can answer that. The difference between 480 and 576 vertical isn't huge, but it's enough...It might be that the framerate acts differently depending on something in the computer. I can't answer that one either. I'm just a vid editor who knows a little about Vegas, but I don't know the hardware coding side of it much at all. I just did the test on a coupla machines. Even my little VAIO was only slightly faster for NTSC than it was for PAL, but nothing as significant as the time you turned in earlier
eheh wrote on 11/30/2002, 6:31 PM
SPOT=DSE (Sorry for that)

Thanks for your help
SonyDennis wrote on 11/30/2002, 9:38 PM
I suspect the extra time is due to scaling of the generated images. In the NTSC case, they are generated at the correct size. In the PAL case, they are generated at NTSC sizes and then scaled to PAL sizes. If there was also NTSC framerate media, it would be doing framerate resampling as well, which would also add time. However, generators are able to generate any framerate needed.
///d@
eheh wrote on 11/30/2002, 10:23 PM
Sonicdennis,
Rendertest by DSE is all generated media.
What you suggest is to start from scratch and insert the generated media on a fresh PAL timeline?
eheh wrote on 11/30/2002, 10:55 PM
SonicDennis,
You were right! Of course!
I found that the generated media tracks had NTSC sizes (720x480) and the pixel aspect ratio was also NTSC's. I simply changed the size to be 720x576 and the aspect ratio to standard PAL and there I have it - 1:47!!!

SPOT, I suggest you have a Rendertest version for us PAL users.
Spot|DSE wrote on 11/30/2002, 10:59 PM
in my version, I was rendering the 720 x 576. That's why my times were so close. I guess I didn't understand when you said you'd converted it to PAL. In the one that you sent me, they were modified. Or maybe I modified them, I'm not sure. Either way, glad it's resolved. If I get enough PAL folks asking, I'll recreate it for PAL. You are the first I'm aware of...
Honestly, the render test was just for folks over at the CreativeCow a few months back, and it was a one-shot deal because everyone was asking about render times. It was a means of establishing a baseline so everyone had the same media...it's not something I really want to spend time at. I only posted the results because a large number of folks asked. I've not been posting results since...
SonyDennis wrote on 12/1/2002, 11:38 AM
Also, if you had separate versions for PAL and NTSC, it would make comparing times difficult. Even if you're a PAL user, render to NTSC to compare times.

That said, be aware that this particular test is really focused on generators, masking, and certain compositing operations. It's kind of like comparing 0 to 60 times for various cars; it gives you something to compare, but there's more to a car than just 0 to 60 mph.

It was nice of Spot to provide a single project that everyone can render and compare their times to the times he has posted. I don't think he needs to worry about adding every single time he receives from this point forward, unless it stands out in some way (beats best time by significant margin, <1:00, first time faster than realtime <g>, etc.).

///d@
Spot|DSE wrote on 12/1/2002, 12:03 PM
The biggest point of this test for me, was to provide something REALLY nasty for the processor to be hit with. Blurs are just plain rough on a render/processor. Same with any generated media, because it's being created, not messing with existing media. Plus, it provides a base line for everyone...I'm only posting some of the newer times if they are impressive. Like the overclocked that came in under 2.00...That was interesting. And of course, the nearly 30 minute PII200 that shouldn't have been able to even do this at all.
salad wrote on 12/1/2002, 12:05 PM
I was hoping to see some times using an HP Pavillion 400-500MHz "Celeron CPU"
I'm going to guess......18 minutes......or a system lock up:)

update:
Thanks SPOT! (LOL)
MarkWWW wrote on 12/2/2002, 5:10 AM
Well, since you asked, here's a result from a machine of a similar spec.

I wasn't going to bother to post the rendertest result from my machine since it's so slow compared to what most people are using, and if I was doing any serious amount of video work I'd have upgraded it a long time ago, but since I use VV3 almost entirely for audio work I'm still quite able to struggle on with this old machine.

Here's the spec:
ASUS P2B-S (Intel 440BX chipset)
Intel Celeron 300A (overclocked to 450MHz)
256MB RAM (PC100)
IBM Deskstar 14GXP 10GB Hard Disk

And it runs SPOT's rendertest in:
11 minutes 21 seconds

Mark
salad wrote on 12/2/2002, 7:22 AM
Mark, that's actually NOT TOO SHABBY!
At least you can OC it!