Creative Communities of the World Forums

The peer to peer support community for media production professionals.

Activity Forums VEGAS Pro Is this lack of difference in CPU vs GPU-enabled render speed a case of the dreaded Kepler chip thing?

  • Stephen Crye

    January 26, 2014 at 3:23 am

    Ok, as per Dave’s suggestion, did some “red car” tests with SVP12.770. These were on my slow T7500, not Ozzie’s faster AMD. In Preferences, Video Tab, tried various renders with GPU acceleration set to either off, or to Nvidia Quadro 2000. Each time I switched it off or on I would restart Vegas as required.

    What is interesting is that the GPU Acceleration setting DOES affect not only preview performance but also the way renders work, with both the Sony AVC and MainConcept AVC encoder.

    As expected, video preview is horrible with it off. In region 5 of the “red car” at Good Half I was getting a measly 5 fps. With GPU Acceleration on, I would get about 19 fps in region 5.

    Sony AVC, with GPU off, but selecting “use GPU if available”, used a bit of GPU, but way, way less than with GPU on. Render time was over 5 minutes; I stopped the test due to boredom.

    Sony AVC, with GPU on and selecting “use GPU if available” used the GPU quite a bit; render time was 3:27.

    Main Concept, with GPU off, rendering using OpenCL used zero GPU. I did not bother letting the test complete – it would have been over 8 minutes.

    Main Concept, with GPU off, rendering using CUDA, was better: 4:47 and the GPU was used a bit.

    Main Concept, with GPU on, rendering using CUDA, give me the fastest render, 3:11 this time ( I hit 3:09 a few days ago).

    Here is screen shot of the GPU graphs for 3 of the tests.

    Steve

    Win7 Pro X64 on Dell T7500, MultiTB SATA, 8GB RAM, nVidia Quadro 2000, Vegas 12, 11, 10, 9 DVDA 6.0 & 5.2(build 135) Sony HDR-CX550V, Panasonic GH3 with LUMIX G X VARIO 12-35mm / F2.8 ASPH, LUMIX G X VARIO 35-100mm / F2.8

  • Dave Haynie

    January 26, 2014 at 3:37 am

    Here’s how to think about this: much of the work done for a preview has to be done for any kind of render. And of course, the resources used to do that stuff, compositing and FX and all, is unavailable for the actual encoding part of the render, when you get around to rendering it.

    So I see about 6x slower renderings going CPU-only than CPU + GPU on this project. That’s not even slightly shocking, given that my CPU benchmarks at about 154 GFLOPs, while the GPU’s peak performance is about 2700GFLOPs. Not the whole story, of course, because it’s very hard to use all of that GPU performance. So I see the CPU-only render pretty nicely pegged at around 95%; when I add the GPU, the CPU use drops to below 25-35% depending on settings, and the GPU use is jumpy, but peaks at 93%.

    -Dave

  • Stephen Crye

    January 26, 2014 at 6:24 am

    [Dave Haynie] ” while the GPU’s peak performance is about 2700GFLOP”

    Yes … and the sad thing is that an nVidia K5000 rated at 2150 GFLOPS goes to waste because of the problems with the drivers and Vegas!

    On a slightly related note, for the 1st time ever I tested the SVP12 “Upload to Youtube” using the red car benchmark. GPU use was zero, zip, nada. The render took about 6 minutes, generated a 95 MB file. I did not time the upload; when I got back to my computer it was already on my channel. Pretty cool actually; all that was needed was to push one button and let the computer do the rest of the work.

    The quality is surprising, and it looks like it automatically took care of the Computer to Studio RGB conversion. Have a look:

    https://www.youtube.com/watch?v=uGMVrMWOiQk

    Steve

    Win7 Pro X64 on Dell T7500, MultiTB SATA, 8GB RAM, nVidia Quadro 2000, Vegas 12, 11, 10, 9 DVDA 6.0 & 5.2(build 135) Sony HDR-CX550V, Panasonic GH3 with LUMIX G X VARIO 12-35mm / F2.8 ASPH, LUMIX G X VARIO 35-100mm / F2.8

Page 2 of 2

We use anonymous cookies to give you the best experience we can.
Our Privacy policy | GDPR Policy