Vegas 11 Build 511 slow render workaround

cold ones wrote on 12/21/2011, 2:03 PM
If you're using Vegas 11 64-bit Build 511 and are getting the agonizingly slow renders, try using the 32-bit Build 510. (I had a simple 5 minute cuts-only timeline: 64-bit render time was 80 minutes, 32-bit render time was under 2 minutes. FWIW, Vegas 10 64-bit also completed in about 5 minutes). In the 64-bit renders, all begins well but a minute or two into the project the output slows to a crawl. 32-bit proceeds at a regular pace. I tried this on two different Windows 7 machines, with & without GPU processing, no difference. YMMV

Instead of reverting back to the previous build, I'm going to continue to edit in Build 511 64-bit and render out of Build 510 32-bit, heaven help me.

Like others, this build has been a near catastrophe for me. I've had the MIDI issue (thanks to those of you who posted about that!), this render issue, and I'm still trying to pinpoint an issue with the iZotope RX2 plugin (I think that using it in a project leads to bizarre audio problems, and greatly increases instability, but I'm weary of gathering more proof.) I've got to get something done now.

Comments

amendegw wrote on 12/21/2011, 2:12 PM
You might try experimenting with "Preview RAM" & "Rendering Threads" see: Render Time Strangeness

...Jerry

System Model:     Alienware M18 R1
System:           Windows 11 Pro
Processor:        13th Gen Intel(R) Core(TM) i9-13980HX, 2200 Mhz, 24 Core(s), 32 Logical Processor(s)

Installed Memory: 64.0 GB
Display Adapter:  NVIDIA GeForce RTX 4090 Laptop GPU (16GB), Nvidia Studio Driver 566.14 Nov 2024
Overclock Off

Display:          1920x1200 240 hertz
Storage (8TB Total):
    OS Drive:       NVMe KIOXIA 4096GB
        Data Drive:     NVMe Samsung SSD 990 PRO 4TB
        Data Drive:     Glyph Blackbox Pro 14TB

Vegas Pro 22 Build 239

Cameras:
Canon R5 Mark II
Canon R3
Sony A9

cold ones wrote on 12/21/2011, 2:56 PM
Jerry, thanks for the tip!

I was able to get the Build 511 to render out in a timely fashion (around 2 minutes) by setting the Dynamic RAM preview max to 500 or less (I'd been going in the other direction, from 1000 to 3000 to 8000, which only made it worse). Changing the Maximum number of rendering threads doesn't seem to help nor hinder (I tried 8 & 16). My Build 510 version had a very low preview max, around 64, perhaps that's what made it work.

Hope this setting is a compromise that will work for both editing and rendering---it would be a pain to change it all the time.

I still have my audio strangeness:
http://www.sonycreativesoftware.com/forums/ShowMessage.asp?MessageID=792324&Replies=0
Any chance this is somehow related to RAM settings?
dxdy wrote on 12/21/2011, 3:25 PM
I have the exact same experience. In 511, I was rendering from 1080i project to NTSC widescreen MPG2 (with sound). With 4000 dynamic RAM, rendering started off well, but crawled to a near-halt, a 2 minute 30 second segment took 18 minutes to render. The only Fx was FBMN's white balance.

I reset the dynamic RAM to 600, restarted Vegas, the same segment with the same Fx took 2 minutes.

I had it set to 4000 (on a machine with 12GB RAM) because that is what was working best the other day. Today, not so much.

i7 950, 12 GB RAM.
amendegw wrote on 12/21/2011, 3:32 PM
Fred, are you saying build 511 is working exactly opposite of what we saw in the Render Time Strangeness thread?

Jeez! I don't what to go thru all those test renders again!

...Jerry

System Model:     Alienware M18 R1
System:           Windows 11 Pro
Processor:        13th Gen Intel(R) Core(TM) i9-13980HX, 2200 Mhz, 24 Core(s), 32 Logical Processor(s)

Installed Memory: 64.0 GB
Display Adapter:  NVIDIA GeForce RTX 4090 Laptop GPU (16GB), Nvidia Studio Driver 566.14 Nov 2024
Overclock Off

Display:          1920x1200 240 hertz
Storage (8TB Total):
    OS Drive:       NVMe KIOXIA 4096GB
        Data Drive:     NVMe Samsung SSD 990 PRO 4TB
        Data Drive:     Glyph Blackbox Pro 14TB

Vegas Pro 22 Build 239

Cameras:
Canon R5 Mark II
Canon R3
Sony A9

dxdy wrote on 12/21/2011, 3:53 PM
Jerry, this is 511, and with the 285.62 Nvidia driver.

And just to contribute to the wierdness...

I tried yesterday's renders again (that convinced me to use the 4000 setting) - no video, just a bunch of jpgs, from various Nikon and Canon point and shoots, all about 2 MB in size, a multi-layer PSD, zooming and panning, with mp3 soundtrack, 41 seconds long.

Rendered to MC MP4 with 600 RAM in 33 seconds. GPU hit 97% (never seen it so high before).

Tried it with 4000 RAM, took 35 seconds, GPU hit mid-80s.

Render parameters:
Audio: 192 Kbps, 48,000 Hz, 16 Bit, Stereo, AAC
Video: 29.970 fps, 1920x1080 Progressive, YUV, 12 Mbps
Pixel Aspect Ratio: 1.000

Running Win 7 64 Home Prem. During the renders, MS Excel 2010 was open as was MS Word 2010.

So I am going to leave Dynamic RAM at 600 until I get frustrated with render speed (again) and start messing with it.

VPro 11.511 64-bit, i7-950 12 GB RAM, Nvidia 560ti 1GB with latest non-beta driver.

Fred
amendegw wrote on 12/21/2011, 4:02 PM
"So I am going to leave Dynamic RAM at 600 until I get frustrated with render speed (again) and start messing with it."Phew! I guess that's all that can be done - I wish there were some consistency here.

...Jerry

System Model:     Alienware M18 R1
System:           Windows 11 Pro
Processor:        13th Gen Intel(R) Core(TM) i9-13980HX, 2200 Mhz, 24 Core(s), 32 Logical Processor(s)

Installed Memory: 64.0 GB
Display Adapter:  NVIDIA GeForce RTX 4090 Laptop GPU (16GB), Nvidia Studio Driver 566.14 Nov 2024
Overclock Off

Display:          1920x1200 240 hertz
Storage (8TB Total):
    OS Drive:       NVMe KIOXIA 4096GB
        Data Drive:     NVMe Samsung SSD 990 PRO 4TB
        Data Drive:     Glyph Blackbox Pro 14TB

Vegas Pro 22 Build 239

Cameras:
Canon R5 Mark II
Canon R3
Sony A9

ritsmer wrote on 12/21/2011, 4:09 PM
coldones wrote: Changing the Maximum number of rendering threads doesn't seem to help nor hinder (I tried 8 & 16)

16 or even 8 are mostly far too high - best result is probably some 2-3.

As to the preview RAM: often there is a sharp peak somewhere between 50 and 90 MB.
cold ones wrote on 12/21/2011, 4:22 PM
ritsmer, I am seeing a nice speed increase if I starve the RAM down to around 50---crazy, considering I had goosed it to 8000 at one point. I still don't see much difference by raising/lowering rendering threads, however. I've dropped it down to 2, 3, & 4 without any difference to speak of. However, the changes in RAM made a most dramatic improvement. You made my day!

One other weird thing: when my render completes, the Elapsed time value reverts to 00:00:02 almost immediately, and if I'm not staring at it I lose the number. I have a second setup where the Elapsed time remains after the conclusion of the render, which is proper. Anyone else see this?

<edit> Nope, I stand corrected, in v11 they both zero out, although the other setup defaults to 00:00:01. Curiouser & curiouser...
dxdy wrote on 12/21/2011, 5:26 PM
After reading the last post or 2, I dropped Dynamic RAM to 60. There was no change to the MP4 times, but the MC MPG2 dropped from 2 minutes to 90 seconds.

GPU usage still jumped into the 90% bracket several times. I believe that the Frederic Bauman white balance plugin I am using is GPU enabled.

Fred
amendegw wrote on 12/21/2011, 5:31 PM
"but the MC MPG2 dropped from 2 minutes to 90 seconds.Fred, I kinda wish you hadn't posted that as I'm 8 hours into a 10 hour mpg2 render. I could have been done 2 hours ago???

...Jerry

System Model:     Alienware M18 R1
System:           Windows 11 Pro
Processor:        13th Gen Intel(R) Core(TM) i9-13980HX, 2200 Mhz, 24 Core(s), 32 Logical Processor(s)

Installed Memory: 64.0 GB
Display Adapter:  NVIDIA GeForce RTX 4090 Laptop GPU (16GB), Nvidia Studio Driver 566.14 Nov 2024
Overclock Off

Display:          1920x1200 240 hertz
Storage (8TB Total):
    OS Drive:       NVMe KIOXIA 4096GB
        Data Drive:     NVMe Samsung SSD 990 PRO 4TB
        Data Drive:     Glyph Blackbox Pro 14TB

Vegas Pro 22 Build 239

Cameras:
Canon R5 Mark II
Canon R3
Sony A9

dxdy wrote on 12/21/2011, 5:39 PM
Jerry, which of your machines are you doing this on? If it is the one with the GT9500, and if the GPU is really important, my 560ti is substantially more powerful. If it were still lollygagging around at 3 - 5% (as it used to do) I wouldn't be paying attention to the difference, but when it is running in the 90% range, that could be a big part of the difference.

Fred
dxdy wrote on 12/21/2011, 5:47 PM
And...with the dynamic RAM set to 60, my preview at Best Half is full 29.97 WITH the Bauman white balance. The GPU is running 22-23% during this preview.

Fred
amendegw wrote on 12/21/2011, 6:21 PM
It's the one with the GT5500 - and CPU was at 100% so, in retrospect, I don't think there was any slowdown.

...Jerry

System Model:     Alienware M18 R1
System:           Windows 11 Pro
Processor:        13th Gen Intel(R) Core(TM) i9-13980HX, 2200 Mhz, 24 Core(s), 32 Logical Processor(s)

Installed Memory: 64.0 GB
Display Adapter:  NVIDIA GeForce RTX 4090 Laptop GPU (16GB), Nvidia Studio Driver 566.14 Nov 2024
Overclock Off

Display:          1920x1200 240 hertz
Storage (8TB Total):
    OS Drive:       NVMe KIOXIA 4096GB
        Data Drive:     NVMe Samsung SSD 990 PRO 4TB
        Data Drive:     Glyph Blackbox Pro 14TB

Vegas Pro 22 Build 239

Cameras:
Canon R5 Mark II
Canon R3
Sony A9

ritsmer wrote on 12/22/2011, 2:43 AM
@cold ones: do you see any change in render speed using your different graphics cards? - or is just the GPU usage percentage lower using the faster card??
cold ones wrote on 12/22/2011, 10:39 AM
Did a quick test, rendering a 5 minute 1080i clip to MP4, RAM=90, THREADS=4

GeForce GTX 450 = 2:12 GPU Usage max ~60%
GeForce GTX 580 = 1:58 GPU Usage max ~40%
No GPU acceleration = 2:07
ritsmer wrote on 12/22/2011, 11:22 AM
Great. Thank you.

Surprising result as the 580 has more than twice the power of the 450 - but then clearly indicating that the bottleneck is somewhere else...

Probably you would get (much?) faster rendering when GPU assist is off if you change the settings to 10 threads and 500 MB Preview RAM ?

Could you please repeat that test rendering to the standard template
"Blu-ray 1920x1080-50i, 25 Mbps video stream" and Best and pls note the CPU usage too ??
cold ones wrote on 12/22/2011, 12:07 PM
Did a quick test, rendering a 5 minute 1920x1080-60i clip to:

Blu-ray 1920x1080-50i, 25 Mbps video stream
(BTW, I didn't have this as a standard template, I had to create it)
RAM=500, THREADS=10
No GPU acceleration = 12:36 CPU Usage max ~95%

--also tried this---
Blu-ray 1920x1080-50i, 25 Mbps video stream
RAM=50, THREADS=4
GeForce GTX 580 = 6:40 GPU Usage max ~55%

In the second try I lowered the RAM & threads based on some of my earlier tests, but I'm afraid that the devil is in the details. We've changed render formats & movie size, etc. My original renders were to much smaller movies than these, 480x270 MP4s. Hope this helps---
Guitartoys wrote on 12/22/2011, 9:45 PM
Good gosh, I wish there was some rhyme or reason to the rendering times with respect to GPU, threads and RAM settings.

I never see my GPUs go above 35%, ever. I am running dual GTX 580's and unfortunately, I have to run SLI to render. So I need to revert to enable all displays for editing. So I can set the 3rd display for my proper preview.

It seems like the Main Concept MPEG-2 is least "tuned" for GPU.

I don't know if I just hadn't noticed before, but the Sony AVC/MVC on the system tab have a Check GPU button. I hadn't noticed this before, maybe it was always there. But these seem to be better tuned for GPU, as my usage is much higher with GPU when using these.

Core/Thread count seems to be project dependent. For example, I blaze through the render test with all of my cores. And reducing cores slows things down.

But when just rendering plan video, with nothing special going on, I can reduce the thread counts to 2 and 3 with virtually no change in render time (+- a couple of secs)

The Dynamic RAM setting seems to have the biggest difference. But If I have 12GB of RAM, I would think that changing between 512MB and 4GB would have little change, but it does.

The one frustrating thing is it seems like I wil fly through the first few minutes of a render, and no matter the setting, things just slow down. I'm running 4 drive SSD RAID 0 and 4 drive RAID 5, and going to either drive makes no change. All I can think is I'm running out of cache on the RAID controller.

Man, it would be really sweet if Sony simply just said, if you are doing primarily X, these are the settings, and if Y, these are the settings. Just some guidance to keep us from burning time trying to figure out how to reduce render time.

It's really frustrating to render for a couple of hours, and then mistakenly change something and have it done in a few minutes. Only to realize you have been wasting hours rendering things due to a simple setting change.

M
Siby wrote on 12/22/2011, 11:46 PM
How do you check the GPU usage in your machine while you render a project?
vkmast wrote on 12/23/2011, 2:05 AM
Siby,
try the free GPU-Z.
ritsmer wrote on 12/23/2011, 2:05 AM
The program from MSI works with most graphics cards: http://event.msi.com/vga/afterburner/download.htm

@cold ones: Thank you. Could you please repeat your second test (the --also tried this---) using the GTS450 (I am trying to find out whether it would pay off to change my GTS450 to a faster card - or if the -still unknown bottleneck- will choke a faster graphics card.

@Guitartoys: agree - would be nice if SCS could give us - just some - hints like "we have tested several different combinations of hardware and it seems that: more memory on the g-card or a PCIe3 interface or lighting a candle and praying while rendering - will bring the fastest render speed when you render to mp4 or mpeg-2 i or p etc" - just some more hints from their no doubt extensive internal knowledge base would (probably) help us getting into the right direction...

And as to the Preview RAM - well - a large preview RAM (some GB) makes it possible to preview time critical parts of our timelines - but just strange that we have to go into Options + Preferences + Video to change this setting to some 50 MB every time we want to render or even do a prerender - and then change all the way back for RAM previewing again...
Here I have made a work-around using a low resoloution template at preview quality and then check the time-critical parts in the Windows media player - and grrrr - now they have removed the Ctrl+Shift+M / the Preview in Player too to make this take several more mouse clicks.
Sometimes I wonder quite a bit who is making these decisions... Sir?
larry-peter wrote on 12/23/2011, 9:45 AM
I'm grateful you all are doing such in-depth testing. I've been able to up my performance a bit in both playback and rendering because of this thread. Thanks.

I'm sure someone has already brought up this obvious question: Why do we have to deal with this at all? The setting is supposedly for "preview RAM!" The fact that timeline performance and rendering performance are affected show either shoddy basic programming or incredibly poor RAM housekeeping.

In other programs that use a memory cache for previewing segments that can't be played in real time - After Effects for example - no other performance is affected (or if it is, memory/GPU cache refreshing is done so transparently you don't notice)

If Vegas is doing this out of some sort of necessity, building upon old code or whatever, there should be some sort of "optimization routine" run either automatically or manually that saves optimum RAM settings for timeline, rendering and preview for your particular machine, switching between them depending on the job being done. In most software development, wouldn't that be listed on the programmer's flow chart right after "load program"? Both Vegas and Sound Forge do a similar type of audio settings optimization based upon audio hardware.

This is the season for wishing, right?

Merry Christmas, all.
Larry
Frederic Baumann wrote on 12/24/2011, 3:23 AM
Hi all,

Fred, if this may clarify the way FBmn Software's GPU accelerated plug-ins work:

- the provided 'GPU Select' utility lets you chose which processing unit to be used: either the CPU or a recognized GPU (must be nVidia/Cuda). You also have an Automatic mode (default) in which case the most powerful compatible GPU will be used. If you change the settings from GPU Select while Vegas is running, it's not an issue. But you will have to restart Vegas so that the change can be taken into account.

- you may check at any time which CPU/GPU is being used, by opening the About box of the plug-in property sheet: it will display the selected processing unit.

- and if you have the non-GPU version of White Balance or Exposure, the GPU will obviously not be used - you may check that in the About box: it will say nothing about the activated processing unit, because these versions just use the CPU.

So far there has been no bug reported on White Balance and Exposure. Plug-ins' GPU usage may range from 95% to 5-10% depending on the power of your GPU, the power of your CPU, and the overall speed of your system (RAM, bus, ...)

Hope this helps,
Frederic - FBmn Software
dxdy wrote on 12/24/2011, 6:07 AM
Frederic, I wasn't complaining. I am delighted with the performance of the white balance GPU plug in, both visually and speed-wise.

Fred