|
|||||||||
|
Thread Tools | Search this Thread |
April 11th, 2007, 03:33 AM | #1 |
Inner Circle
Join Date: Feb 2007
Location: Tucson AZ
Posts: 2,211
|
Multi Core / Multi Processor Systems
Does Vegas (or any other similar package) take advantge of multi core and/or multi processor systems if they're present?
I'm familiar with programs like Cinema 4D which will spread rendering across multiple processors or spread frame ranges across multiple networked systems and render frames in parallel and wondered if Vegas would do something similar. |
April 11th, 2007, 04:25 AM | #2 |
Trustee
Join Date: Nov 2005
Location: Honolulu, HI
Posts: 1,961
|
Yes. I've heard up to four. I know at least two work, and a fast dual-core can edit native .m2t files without going to an intermediate with Vegas 7.
|
April 11th, 2007, 09:24 AM | #3 |
Major Player
Join Date: Aug 2006
Location: Glasgow/Scotland
Posts: 626
|
John Rofrano is using a Quad Core and says it makes a difference when rendering. Reading reviews about the Quad Core all say that video editing apps are one of the main applications to benifit.
I'm anxiously awaiting the 22nd of April when Intel are due to reduce Quad Core prices by 40%. It's still going to be expensive but.......I'd say this is one area it's worth investing a bit extra in on Vegas which relies/makes more use of the cpu than most. |
April 11th, 2007, 07:01 PM | #4 |
Inner Circle
Join Date: Feb 2007
Location: Tucson AZ
Posts: 2,211
|
Thanks for all the responses. This is good news. I'm looking at getting an 8 core system (2 quad cores) with 4GB in the next few weeks.
Main reason is the render times for CG animations where they have to calculate light interaction, particles etc. I'm also thinking of adding some computational fluid dynamics to the witches brew of applications. I suspect that rendering in Vegas isn't as computationally intensive as CG and fluids rendering, but I was thinking/hoping that it might also benefit from the added parallelism. (What seems like 100 years ago I did some work on a supercomputing project aimed at the non-parallel application market) |
April 11th, 2007, 07:17 PM | #5 | |
Inner Circle
Join Date: May 2005
Location: Windsor, ON Canada
Posts: 2,770
|
Quote:
I built my new "beast" based on his specs and the results were phenomenal!! The VASST render test was done in 14 sec. My personal "render test from hell" was a project from last summer. It was a 10 minute video that was heavy on chroma key stacked with assorted titles, graphics and other FX. A work machine (P4 3.4GHz) took 3 hours :-( My beast did it in 27 minutes :-) Other examples: 1 hr. (stills) to NTSC DV-AVI: 18 min. 1 hr. (stills) to MPEG-2 (DVDA NTSC video stream template): 11 min. 1 hr. (video) to NTSC DV-AVI: 33 min. Fast enough for you? > huge grin here < |
|
April 11th, 2007, 08:10 PM | #6 |
Major Player
Join Date: May 2004
Location: Orlando, FL
Posts: 204
|
What is your beast machine?
On the 22nd when the new 4mb Cache, 10x multiplier C2D Intels Come out I'm going to build a dual core box from the ground up. The 1.8 and 2ghz chips can overclock on stock cooling and small if any voltage adjustments to 3-3.3 ghz. People on water cooling have the C2D's up to 5ghz! |
April 11th, 2007, 08:36 PM | #7 |
Major Player
Join Date: May 2004
Location: Orlando, FL
Posts: 204
|
What is your beast machine?
On the 22nd when the new 4mb Cache, 10x multiplier C2D Intels Come out I'm going to build a dual core box from the ground up. The 1.8 and 2ghz chips can overclock on stock cooling and small if any voltage adjustments to 3-3.3 ghz. People on water cooling have the C2D's up to 5ghz! |
April 11th, 2007, 11:33 PM | #8 |
Major Player
Join Date: Aug 2006
Location: Glasgow/Scotland
Posts: 626
|
|
April 12th, 2007, 12:11 AM | #9 |
Major Player
Join Date: Feb 2007
Location: Huntington, WV
Posts: 260
|
multiprocessor question
I have another computer that I considered putting with the one I use now to take advantage of multiprocessing, but, I thought I read that some of the vegas effects and many plugins including magic bullet, is not supported in multiprocessor rendering. Is this true? If so, will it just slow the rendering down or not work at all? And, how do you set the systems up using two computers? J
|
April 12th, 2007, 04:00 AM | #10 |
Inner Circle
Join Date: Feb 2007
Location: Tucson AZ
Posts: 2,211
|
I can't speak for how Vegas does it because until a few hours ago I didn't know that Vegas did it.
However... The way most network render systems (for CG etc!) work is to have a server and several client machines. The server can be any machine on the network. When it's render time the server sends a bunch of frames to each client on the network and the clients render their little hearts out and send frames back to the server when complete. The server keeps track of how many frames have been finished by each client and re-divides up the remaining frames across the clients so in the end the faster clients get more frames. Some applications are clever enough to break up individual frames into segments and send each client a segment, but there are some limits to this. In the case of a multi core - multi CPU in the same box, it works in a similar fashion but communication is better. |
April 12th, 2007, 11:09 AM | #11 |
Regular Crew
Join Date: Sep 2006
Location: Chicago, IL, USA
Posts: 93
|
You'll be happy.
My 8-core machine (dual Clovertown 2.13GHz) outruns my 4-core Intel (Kentsfield 2.66GHz) and 4-core AMD (dual Opteron dual-core 2.6GHz) doing most things with Vegas.
Why would this be when you can only set the video rendering threads to 4? There are multiple video processing threads. There are multiple video reading/decompression threads. There are multiple audio processing threads. There are multiple audio reading/decompression threads. There are compression threads. There are external display threads. etc. On top of this, some sections of the code use multiple threads launched (not exactly, but this isn't a programming forum) from one thread. Basically, Vegas has been eagerly waiting for the multi-core revolution for some time. Now, assuming that the machine that you are getting is a Clovertown (because Barcelona isn't available yet), you have a couple of fast buses. Still, you don't have the world's fastest memory controller. While you'll see performance scaling in most cases, some things will be limited by the memory subsystem. Other things may be limited by disk performance, etc. This is not just in Vegas, but in a number of places. Don't expect things to be 8x as fast. I've seen 4x-6x in quite a few instances, though. Have fun with the machine. |
April 12th, 2007, 12:46 PM | #12 |
Inner Circle
Join Date: Feb 2007
Location: Tucson AZ
Posts: 2,211
|
4 to 6 X would be great. I also have a couple of other reasonably fast machines (dual core) on the network so for CG rendering I will probably have 10 to 12 cores going in total (I did a simple flame and smoke simulation and it took 4 hours to render 43 seconds of 640 x 480 on a single processor a year or so ago)
So CG rendering should be fine, and it looks like Vegas will also be "very happy" I did an early commercial multi threading applications for IBM in 65 when I designed a very high speed fortran compiler for them. Multi threading rocks!!!!! |
April 13th, 2007, 11:44 AM | #13 |
Major Player
Join Date: May 2005
Location: Southport - UK
Posts: 208
|
Mike
Can you run the VASST HD render test? I'm running 7.0d on a Core 2 Extreme QX6700 on XP Home. But when i try that test to 1080-60i it always crashes at the motion blur section betwen 21 and 24 seconds. If I just try to render that section it also crashes. If i render to standard mpeg2 it works fine. Ian |
April 13th, 2007, 11:56 AM | #14 |
Inner Circle
Join Date: May 2005
Location: Windsor, ON Canada
Posts: 2,770
|
Ian, I'll do it sometime this weekend and post my result - or lack thereof :-)
One question though. The VASST site at http://www.vasst.com/?v=render_test_2006.htm shows 2 different tests, old and new. Which one do you want me to run? Mike |
April 13th, 2007, 12:20 PM | #15 |
Major Player
Join Date: May 2005
Location: Southport - UK
Posts: 208
|
Thanks Mike - I appreciate it.
It's "the new render test" - it links direct to the .veg. As it says on there, try to render to 1080-60i. Also, my set up renders the original test in 14 seconds also - but slight overclocking gets that down to 11 :) Ian |
| ||||||
|
|