EEVblog #698 - GPU Video Rendering

preview_player
Показать описание
Dave tests a really high end NVIDIA GTX-970 video card for accelerated CUDA GPU video rendering with Sony Movie Studio (Vegas) & also Adobe Premiere.
Will it work for the new high frame rate video ?
Will it work at all?
How about a Radeon HD7850 and OpenCL?
Does the CODEC matter ?
Does hard drive or SSD performance matter ?
This also provides some insight into how Dave edits and renders his videos.

NOTE: This is mostly trying out and real time commentary of testing video rendering speed. If you find this stuff boring then DON'T WATCH IT.
There is no electronics content here, but plenty of main channel viewers might find it interesting.

Support the EEVblog through Patreon!

EEVblog Amazon Store (Dave gets a cut):
Donations:
Projects:
Electronics Info Wiki:
Рекомендации по теме
Комментарии
Автор

Hi Dave, great video.  I do video editing for a number of projects and what i've found is GPU acceleration only applies and accelerates video rendering in a few different circumstances.  If your video is effects heavy, has lots of animations, or color grading then the parallel processors and frame buffer in the GPU can render the frames much, much faster than any CPU due to the enormous number of graphics cores available.  

However, in a rendering job with primarily just cuts and the source video with few adjustments, the CPU's individual cores are enormously faster than an individual processor unit in the GPU.  The GTX 970 for example has 1664 CUDA cores that can divide up a complex, highly graphics-intensive sequence into a parallel render job, much like rendering objects and textures in video games (it is a gaming card after all).  Conversely, the 4 (8 virtual with hyperthreading) processors in your CPU can shred through unmodified frames very quickly in a serial rendering of your video clips since each core is much more powerful.  Adding more objects and complexity to the render sequence will show more benefit from having parallel processing cores.

ethanperrin
Автор

I ran into the exact same problem when I built my new rig with a GTX 970 and Core i7 4790k.  I use Powerdirector though.  What I ended up doing was upgrading to Powerdirector 13, and I encode with the H264 codec in MP4 format.  1080p 30fps at 28mbps encodes much faster than real-time when using GPU acceleration.

Maxxarcade
Автор

13:20 Thats why you dont buy a gaming card for video rendering...

neclovek
Автор

Dave, your video pipeline is something like:
decoding -> rendering -> encoding
It is normal that you are not getting any speed improvements by using the new GPU because in your case the rendering process only takes a small fraction of time from the total video pipeline! To accelerate your work, you should get a codec capable of using the GPU for encoding and decoding, not rendering. Rendering is important if you have a lot of effects, transitions or overlays in your video.

logiside
Автор

I'm actually very interested in how you would make videos. What you mentioned with all the different formats, as well as handbrake seem, to me, to be very important in making videos. I just have to be patient when working with my videos, as I'm only using a laptop. Too bad that the 970 didn't work for you.

LazerLord
Автор

Dave, can you please provide some sample footage at 50/60fps? I can test the rendering with Vegas 14 with a GTX 680.

AttilaSVK
Автор

This kind of leaves you wondering, what would a top of the line (or near top) openCL GPU would do...

billysgeo
Автор

Thanks for the video Dave, I just encountered this issue. I recently bought a 980ti card for gaming and picked up sony studio 13 and spent the whole evening trying to get it to work with no luck then I found your video and your mirroring my experience exactly.

Haru
Автор

Of course the HDD matters, for example: you have a high end CPU and a high end GPU, and of course the crappiest hdd you can put in there. As long as you don't export low video quality, the hdd will slow your PC down by a significant percentage, as it can't buffer all the video in the RAM

alexzirnea
Автор

And yes, the codec also matters because different encoders have different compression levels and require different processing power

alexzirnea
Автор

I don't understand why you save it as another format, and then convert it by handbrake? I always render as Sony AVC mp4 format and get decent quality vs. size. I just tried you method to render as AVI and then convert with handbrake to mp4 - and ended up with something just as good as the file I got from Vegas. 

SteenPedersen
Автор

Hi Dave,
if you are a noob in GPU aided Video editing, there are different calculating tasks while editing for the PC hardware :
1: decoding the video (can be done in software  (cpu) or the dedicated hardware decoder in a gpu (missing in some highend GFX cards, especially Geforce GTX))
2: editing the video ( titles, animations, audio etc...(might be accelerated by a gpu))
3: encoding the video ( changing the codec / resolution / framerate (might be accelerated by a gpu))
in the video, none of these tasks are performed. Only the IBP Frames where regrouped, which is pure (HDD/SSD) IO Performance of your system.
Monitor the CPU with the task manager on a second screen and the GPU with e.g. EVGA PrecisionX how much load is on the hardware units.
Also keep in mind that you are using a cpu with multiple cores with AVX extensions, especially optimised for video transcoding with highly optimised libraries from intel. Most video editing software use AVX if detected and the benefit of gpu acceleration is much smaller.
cuda has some serious compatibility problems in video editing software. Your cuda hardware must match the supported cuda Version in hardware and software of your video editing software (might not be a technical problem, but a "upgrade to our latest software version for a little fee to pay our programmers" problem).

dasrotemopped
Автор

Did you change the cuda_supported_cards.txt in the program files folder for Pr?
Unless your card is not in there by default, it won't even touch your video card for rendering.
You might also need to go to Project > Project Settings > General and enable Mercury Playback Engine (CUDA)

swc
Автор

Is it not best to render and upload in Quicktime format. YouTube does not then need to second render it. To my understanding?

TheDutyPaid
Автор

What makes you think you need 60fps?? That's only used for slow motion, display video only needs 30fps.

Flintsmooth
Автор

Dave, i use TMPGenc for video rendering and with this CUDA works. The point is, a GPU can only work with static information, MP4 is not static, it's dynamic. If you record your movie with your camera, with a static codec (ex. mpeg 2) the GPU can decode the information and the CPU can encode it into MP4. With this your CPU only have to do one thing, giving a huge boost. You can try to get MP4 to work static if you use CBR, and all other settings to be divided by 2.

Dippo
Автор

Dave, have you updated Movie Studio? The latest for Movie Studio 13 is Build 943. This looks like a issue with the codec/Movie Studio considering the GTX 970/980 are new with a pretty big architecture change. 

shadow
Автор

Dave, have you tried rendering with Quick Sync? It's supposed to be even faster than GPU accel

tylerwatt
Автор

If you are using handbrake for the final encoding anyway, why not go with a simple uncompressed codec for your render?

Or is uncompressed even slower than the compressed one, or is the one you use uncompressed anyway... Also adding in the time to load up handbrake and fiddle with the settings might have some time added/lost there with exporting to a compressed file format from the get go.

I don't know, and I don't claim to be an expert or anything like that. I'm just tossing out questions and maybe I can help see around the 'problem'.

spokehedz
Автор

What happened to the Windows glass feature? Your computer looks so depressed without it.

Browningate