SVP 13 Render Times, Too Good To Be True?

bhurst wrote on 5/1/2014, 12:36 PM
I just started playing with SVP 13 and am amazed at the render times. Rendering the same project, same codec, same machine ... same everything ... the render time is about 75% less compared to SVP 12.

This is a fairly complex project although short ... PIP, overlays, color correction, etc. ... rendered to 1080/30p.

What takes about 12 minutes in SVP 12 takes about 3 minutes in SVP 13.

Too good to be true? Some yet-to-be-noticed tradeoff? Or just a near-miraculous "discovery"?

Comments

videoITguy wrote on 5/1/2014, 3:18 PM
Your finding might be miraculous, although I wonder....your generalizations does nobody any good here - you left out the details, and details, and and more details.
bhurst wrote on 5/1/2014, 3:41 PM
You're being just a bit rude for no reason I can think of. That fact that the exact same project rendered with the exact same settings was four times faster with v13 than with v12 is significant.

Actually, I rendered several different short projects using both versions with about the same ratio of improvement with v13.

I even used the rendertest-hdv.veg test project that has been used as a benchmark elsewhere on this forum. Using 1080/30p-MPG2 output the results for that benchmark were 26sec/v12 and 4sec/v13.

That was using my system #2 in my profile.
Red Prince wrote on 5/1/2014, 3:50 PM
your generalizations does nobody any good here

That was uncalled for.

He who knows does not speak; he who speaks does not know.
                    — Lao Tze in Tao Te Ching

Can you imagine the silence if everyone only said what he knows?
                    — Karel Čapek (The guy who gave us the word “robot” in R.U.R.)

DavidMcKnight wrote on 5/1/2014, 3:57 PM
To accurately test, delete any rendered file, reboot the PC, restart the other version of Vegas and re-render.
John_Cline wrote on 5/1/2014, 4:25 PM
Using my original "rendertest-hdv.veg" file and rendering in VP13 to the stock "HDV 1080-60i" MPEG2 template with video rendering quality set to "Best", my i7-920 quad-core laptop with a Quadro FX1800M video card rendered in 1:54 with GPU video processing turned off and :30 with GPU turned on. This is certainly consistent with the 4X speed increase you are seeing.

Based on this result, I'm surmising that you don't have GPU video processing turned on in VP12 or there is some new GPU code in VP13 that works better with your video card.

(I just ran the same test in VP12, GPU-on=:35, GPU off=1:56, so it seems that VP13 is just a hair faster than VP12 on this particular test and a good 5 seconds faster with GPU on.)
videoITguy wrote on 5/1/2014, 6:52 PM
Hmm, seems John Cline chose to spoke with clear and complete details - this is just the kind of information that needs to be shared with the user community and gives an informative picture of what we all need to be aware of. Thanks John!
wwaag wrote on 5/1/2014, 7:23 PM
Using the same "rendertest-hdv" and same template, I likewise don't see any real differences between V12 and V13. Here are my numbers:

V12: Nvidia GPU 16 sec, Intel GPU 20 sec, CPU only 45 sec.
V13: Nvidia GPU 16 sec, Intel GPU 19 sec, CPU only 51 sec.

wwaag

AKA the HappyOtter at https://tools4vegas.com/. System 1: Intel i7-8700k with HD 630 graphics plus an Nvidia 1050ti graphics card. System 2: Intel i7-3770k with HD 4000 graphics plus an AMD RX550 graphics card. System 3: Laptop. Dell Inspiron Plus 16. Intel i7-11800H, Intel Graphics. Current cameras include Panasonic FZ2500, GoPro Hero11 and Hero8 Black plus a myriad of smartPhone, pocket cameras, video cameras and film cameras going back to the original Nikon S.

Stringer wrote on 5/1/2014, 7:26 PM
Re. OP,

The only answer that occurs to me is that V12 was/is somehow crippled for you.

If V13 was 4x faster, or even 2x faster, Sony would be crowing about it from the rooftops, and many other users here would be shouting it's praises..
John_Cline wrote on 5/2/2014, 12:37 AM
What is evident is that GPU video processing works and can significantly speed things up. Track 3 of my "rendertest-hdv.veg" uses Sony Gaussian Blur which is one of the plugins that is GPU-enabled. For anyone else that wants to try it, it can be downloaded using the following link:

http://www.johncline.com/rendertest-hdv.veg

I updated it in 2010 and that test .VEG can be downloaded at this link:

http://www.johncline.com/rendertest-2010.zip

This .VEG should take four times longer to render than the original rendertest-hdv.veg file.
A-Scott wrote on 5/2/2014, 1:26 AM
Sounds like multi-core situation. Is your SVP 12 using all your CPU cores during render?

bhurst wrote on 5/2/2014, 5:23 AM
All my tests were with GPU on for v12 and v13. As I said before, all settings were the same.

Surprisingly, the average CPU usage by v12 was considerably higher than by v13 even though v13 was so much faster. All cores were active at different levels during all tests. Memory use was about the same for both versions.

John_Cline wrote on 5/2/2014, 6:03 AM
Are you using System #1 or System #2 listed in your profile?
bhurst wrote on 5/2/2014, 7:50 AM
System #2 (work PC).

I did rerun the rendertest-hdv test this morning and got different results.

v12, GPU on, stock 1080/60i-MPEG2, Best, 10 sec
v13, GPU on, stock 1080/60i-MPEG2, Best, 8 sec

Not sure what I did different, but my apologies for any erroneous data. I knew something was not right, but ...

In any case, this is the best results I've ever gotten using your benchmark. The only difference is this is the first time I've run the test since I replaced my old 5000 series AMD graphics with a new 8000 series.

I just got the new home system (#1). I haven't even started installing software on it yet, but it should be a real "barn burner."
DanH wrote on 5/2/2014, 7:59 AM
Is your home system custom made? They look like nice specs.

Dan
bhurst wrote on 5/2/2014, 8:03 AM
I guess you would say custom made. Put together by CyberPowerPC (http://www.cyberpowerpc.com/).

My previous home PC came from them, and it's worked well for five years, but I got the urge!
bhurst wrote on 5/2/2014, 8:09 AM
John,

I ran your new, 2010 test with the following results.

v12, GPU on, stock 1080/60i-MPEG2, Best, 31 sec
v13, GPU on, stock 1080/60i-MPEG2, Best, 27 sec

Thanks for providing these benchmark tools!!

B
set wrote on 5/2/2014, 8:42 AM
Bhurst, perhaps you may want to try other test like this previous thread, by JohnnyRoy...
bhurst wrote on 5/2/2014, 9:05 AM
Thanks, "set".
Rv6tc wrote on 5/2/2014, 10:08 AM
Mine was eye-opening.

GPU on:
VP11 0:44
VP13 0:53

GPU off:
VP11 2:55
VP13 3:41


I tried to post this to Steve Mann's site, but got multiple errors when I went to enter it.

How can VP 13 be that much slower?
dxdy wrote on 5/2/2014, 10:28 AM
What were you rendering from? What were you rendering to?

(One post with two sentences ending in prepositions. Yikes)
ritsmer wrote on 5/2/2014, 10:35 AM
"How can VP 13 be that much slower?"

Yes it is strange - but precisely what I found here on my machine too.
It was with my normal work formats: input full HD AVCHD and output full HD MainConcept m2t.

V13 did render some 10% slower - V13 test-version, with watermark, that is, but I doubt that the watermark takes so much extra time.
Rv6tc wrote on 5/2/2014, 12:52 PM
dxdy,

It was John Cline's speed test. Rendered to HDV, MPEG-2.

Thanks, ritsmer. At least it's not just me.....
NCARalph wrote on 5/8/2014, 7:01 PM
With John Cline's 2010 speed test and NVidea 570 GTX GPU:

1080i
V11: 44 sec
V13: 50 sec

1080p
V11: 15 sec
v13: 17 sec

CPU rendering:
1080i
v11: 6:03
v13: 5:59

It looks like on my system the two versions are about the same on just CPU but V13 is substantially slower on the GPU than V11. I've seen the same kind of results on my own videos as well, so it's not just that test file.

I'm using build 290 and am downloading 310 and will report back. There is no change with 310, the rendering time is the same.
bhurst wrote on 6/3/2014, 7:04 AM
I finally got around to running the original test on my new PC ... System #1.

v13, GPU on, stock 1080/60i-MPEG2, Best, 7 sec

If I customize the codec for 30p, which is what I actually use, it is 3 sec.

However, I find that with my "real-world" projects, I often get faster renders with the GPU off ... not sure why.

B