GPU rendering in Vegas Pro 13 - no improvement

akm wrote on 10/6/2014, 9:08 AM

I just upgraded to Vegas Pro 13 last week. Likewise, I upgraded my video card to a Radeon HD 6970 based on other user testimony that this card in particular works well with the GPU rendering aspect of Vegas. However, after the upgrade, I'm not noticing a difference at all in the rendering speeds, and actually, it may even be taking longer now (for instance, a 10 minute long video with one video track and one audio track is taking taking 45+ minutes to render, with the GPU rendering on; 1080p, 30 fps). I have attempted testing it with varying file types based on older Sony guides (.mp4, .avc, etc), but no improvement there either. I have ensured that I have the latest graphics drivers and I have the latest build of Vegas Pro 13. GPU rendering is also enabled in the Preferences tab. Is there anything else I may be doing wrong? It won't be the end of the world if I can't get it fixed (the card is already working wonders for my games), but the main point of me having bought it was to improve the performance of Vegas and in particular my render times.

Comments

OldSmoke wrote on 10/6/2014, 9:30 AM

When you render using the MC AVC templates you must select OpenCL, don't leave it in automatic. For the Sony AVC template there is only CPU, GPU or QuickSync, select GPU.

Aside from all that, eventually your CPU will be a bottleneck and the GPU isn't much of help anymore. The CPU still needs to help prepare the frame and pass it to the GPU.

Also try and use a monitoring software, HWinfo64 or similar, to see how much GPU load you get during playback and rendering.

akm wrote on 10/6/2014, 1:24 PM

Thank you for the response, I will give that a try in regards to the monitoring software.

One issue I stumbled upon that I corrected after my initial post is that I wasn't configuring the exports for GPU rendering in the AVC and MP4 custom-config screens. I did that and ran a test, but it still wasn't ideal-- a 20 minute long video still took an hour to export.

I'm beginning to wonder if my main hardware is acting as a bottleneck (older AMD Phenom II quad core CPU, 3.0 ghz, with DDR2 800 RAM, 8 gb). Hm. I've heard of excellent results with this card, and my personal experience definitely isn't matching up with others' experiences. However, a lot of them likely had i7 rigs which are known for being workhorses anyway.

**edit (about an hour later): I am currently exporting a video using MC AVC. I set its config to use Open CL, and exported with 1080p, 30 fps. It's a 31 minute long video, and it's looking like it will be done after 38 minutes. So, that's a much better result (we'll see how the quality of the actual file is afterwards though). Still not the improvement I hear people talking about, but it really could just be the rest of my hardware acting as a bottleneck. I loaded up HWinfo64 and it seems like my GPU is sitting at a solid 64% utilization rate during the export, so it's definitely being used.

astar wrote on 10/6/2014, 2:18 PM

Run GPU-Z and verify the card is operating at 16x speed, some motherboard steal pci lanes for other high bandwidth devices. You may need to adjust device enablement or remove something is GPU-z shows the device running at 8X. If your GPU is does not have full bandwidth to card and memory, this may bottleneck bidirectional communication speed or increase latency timings with the application and hardware.

Run winsat memformal from an admin console. Your memory bandwidth should exceed 8GB/s.

If all that is golden, then that CPU motherboard might be holding you back. Try CPU-z and verify the instruction set is similar to the i7. There might be an instruction set that is missing, that may be holding back the math used to speed the compression. You may also be seeing the lack of hyperthreading, and how much faster i7 is at the math associated with compression.

My i7 with an hd5770 will push the cpu to about 50% and the GPU to about 75-85% using Sony's AVC codec. Playback the GPU will hit about 20% to 65% depending on the codec and layers of effects. You can use AMDs system monitor to see the CPU and GPU utilization in real time.

akm wrote on 10/6/2014, 2:55 PM

Great, thanks for the tips! I'll perform those checks later tonight.

akm wrote on 10/11/2014, 7:29 PM

I haven't run those tests yet, but for those curious, I have found that export speeds are vastly improved as long as I select a AVC or MP4 export option that specifically states "Use OpenCL when available". If I use one that just says "Use GPU when available", it doesn't seem to render with the GPU. Also, I have found that certain video containers work better and faster with the export process than others. For instance, if I have a project built with WMV files, the export process is extremely slow (for instance, right now I'm rendering a 20 minute video and it is taking an hour to do so; all WMV clips are being used as the source files). However, if I have a project that uses source files Vegas seems to "like" more, then when I export to a MP4 or AVC file (using the correct template as mentioned above), then it exports very quickly--just about a minute per minute of footage (so, a 20 minute 1080p video takes about 20 minutes to render). This may not be as fast as it could be (re: other bottlenecks), but it's a massive improvement over my old video card and Vegas Pro 10 which I was using until a couple of weeks ago.

Thanks for the help everyone! I think my problem is solved. :)