OldSmoke:

Here goes:

I used the vp11 test/demo project (the one with the red Mercedes). I hope that's the project you meant. All tests were with VP12, Build 770.

I TRIED to keep everything as constant as possible, but my results are so inconsistent, that I'm truly befuddled. Given that I was switching cards between machines (requiring restarts), I can't say that I was totally successful.

I report the mode (i.e., most often observed) and then the range in parentheses, when I tried multiple renders.

System 1: Intel i7 980 3.3GHz

MC AVC Internet 1080p Template
GPU Off: 167 secs
GTX 560 Ti: 62 secs (62-93 secs)
GTX 770: 84 secs

XDCAM EX HQ 1920x1080-60i 35 Mbps VBR Template
GPU Off: 195 secs
GTX 560 Ti: 78 secs (78-82 secs)
GTX 770: 55 secs


System 2: 2x Intel Xeon E5-2620

MC AVC Internet 1080p Template
GPU Off: 175 secs
GTX 560 Ti: 239 secs
GTX 770: 133 secs (114-135 secs)
GTX 780: 114 secs (104-135 secs)

XDCAM EX HQ 1920x1080-60i 35 Mbps VBR Template
GPU Off: 247 secs
GTX 560 Ti: 234 secs
GTX 770: 77 secs (77-77 secs)
GTX 780: 72 secs (72-84 secs)

I would be very mad if I spent all the money trying out vary video cards and got those times as JasonATL got.

AMD 6970 here... 36s GPU ON 103s GPU off. Sony Vegas Benchmark Project testing.

Bruce - I'm not sure I understand your point and certainly have no context for your times, given the lack of specifics on what the times refer to.

It seems that the speed increase that you achieve is comparable to that of the GTX 770: GPU render is approximately a third of the time of the non-GPU render, at least for the XDCAM render. What is there to be mad about? The MC AVC gain isn't nearly that much. Plus, you have a very fast Vegas system to begin with, apparently.

Having said that, my real complaint about VP12 is its stability, especially related to GPU rendering. It didn't crash on these tests, but does on most of the projects that I do (involving DNxHD files). GPU acceleration seems to exacerbate the problem, though is by no means the only problem. Frustrated? Yes. Mad? Well, I use PPro and Resolve proportionally more and more every day.

More to the Grazie's question, my results probably don't appear to suggest that more CUDA cores means a proportional gain in speed. More does appear better, but comparing a 560 Ti and 770 or 780 is somewhat apples to oranges.

I'll also add that, unlike Resolve, Vegas does not use two cards. Resolve makes use of both cards, effectively doubling the speed of my system. PPro also uses both cards, though I have not benchmarked it to determine whether the speed doubles with two (I suspect not).

That, in essence, should be the first hint at answering Grazie's question. If Vegas liked more CUDA cores, like Resolve does, it should add cards together, not give you the choice of which to activate.

Jason,

I quote..
I used the vp11 test/demo project (the one with the red Mercedes. I believed you are talking about the Sony vegas benchmark project.. Yes, I would be very mad if I spent $500+ on the 7xxx series cards and gotten those terrible gpu rendering performance. Base on the information you have post up here. You can't be telling me that you are happy with those times. Are you?

Hi Jason. Thanks for taking the time and posting the results. Sorry for getting bashed by Bruce, that is why Grazie wanted you to do it off the forum. Anyhow, I do believe that there is something wrong in your system aside from GPU acceleration. Have a look at Hulk's benchmark study where many users have submitted the results of their systems. An i7 980 isn't a slow cocker by any means and your dual Xeon isn't either. For both systems I would have expected a bit shorter render times without GPU then you got from your test. So here are a few questions just to make sure we compare apple to apple.
The SCS benchmark project properties must not be altered, render quality should be set to Good, it should be 8-bit too and not 32. What are your preview ram settings? If possible leave it at the default. Set the window layout to the default; doesn't make much difference but it does depending on your system. When you render to MC AVC ensure you have selected CUDA in the template, don't leave it on automatic. You may also have a lot of services running in the background that can cause longer render times. How are your drives setup?

I do run 2x GTX580 on my system and Vegas does use both of them. Try selecting the card that is in the lower slot for GPU acceleration; disable SLI or set to activate all displays. Driver is another issue. I do know however that if you want to use GPU acceleration in PP and Resolve you need one of the newer 300 series drivers. Those drivers will slow you down in Vegas when using a 5xx card, the fastest is 296.10 which doesn't support the 7xx cards. On my system the difference between the 296.10 and latest 331.?? Driver is 4sec., 29sec vs 33sec for MC AVC rendering.

There is a big but too. If you mostly work with DNxHD files and also render to that format, GPU acceleration with Nvidia will not be as good as with formats outside QuickTime.

JasonATL,

"MC AVC Internet 1080p Template"

Did you enable GPU use in the MC AVC encoder or leave it at CPU only Specifically did you select, "Render using CUDA if available", in the template options.

Only with this option will you get the most out of your GPU during a "render as". This option allows the encoder to use GPU.

AMD 6970 here... 36s GPU ON 103s GPU off. Sony Vegas Benchmark Project testing.


Same here. 36 seconds here using XDCAM EX 1080-60i template.
AMD 7950, i7 4770 CPU.

Oldsmoke,

My goal here was to help Grazie and others. I am not happy with Vegas results, as I think I have said. I appreciate your suggestions for how to improve the results. If I cared about Vegas performance on my system, I might try them. But, my results with Resolve and PPro are in line with benchmarks on similar systems. Since that is where I spend my productive time, I have little interest in using drivers that work with Vegas, for example, or spending time tuning for Vegas, which I use less and less due to the effort you describe in getting it to work well.

To answer your questions, I don't think I altered the project in any way. Window layout is most certainly not default and ram preview is 200. I chose use cuda for the gpu results. Not many services should be running on sys 2. Not even virus protection. Whatever services are running are important to me for the work I do with Resolve or PPro. It does have recent nvidia drivers, for the reasons you point out.

Perhaps Vegas uses both cards for you since they are identical? Resolve and PPro use both regardless, which is not to say they don't care.

Bruce's comments are fine by me. The point of my response here was that I thought I could contribute to answering Grazie's question (Grazie has certainly been generous and patient in helping me through the years), since I have these cards and, while perhaps settings are not optimized for Vegas, they are held constant in comparing these cards. Take them for what they are, not for what they could be.