Dedicated GPU or Native CPU for Nuendo Video Playback?

Last question for a while :slight_smile: Couldn’t find a straight answer in any of the forums related to this.

Is it worth looking into using a dedicated GPU for Nuendo? I’m pricing out a new PC build and GPU is the last piece to plan out. Does anyone use dedicated GPU card for video playback, or is just leaving it on CPU, or even a blackmagic decklink mini, better? If using a dedicated card/blackmagic are there any latency or delay issues to be worried about?

I know that video file format is the first thing to get right. But is there any significant CPU performance or is it marginal?

If it helps this is the rest of the build:

In a nutshell: dedicated GPUs are for gamers or 3D CAD pros.

For basic video editing, the built-in GPU on recent CPUs is more than 99% will ever need.

The remaining 1% should try an entry-level (pref. fanless) GPU before wasting money on fancy high-end (and often noisy) GPUs.

2 Likes

To expand on that, for video playback of any common format (like AVC, HEVC, VP9, etc) graphics chips, both integrated and discrete, use special dedicated hardware. They have ASICs onboard designed just to play back that kind of video while using minimal power. As such, any of them are going to do what you want. The decoding and playback is handled by a special part of the chip and doesn’t really put load on anything else.

Where it starts to matter, for video at least, is if you are doing either a whole lot of decoding of different streams at once, doing encoding, or doing effects/compositing. Then a dedicated card can matter.

It matters the most for effects/compositing since those run on the rendering part of the card itself, if the software you use knows how, and thus can run much, much faster. Encoding can also be something they are better about, like Intel CPUs know how to encode AVC, but they don’t do nearly as good a job as nVidia does, so if you are using the accelerator to do encode, rather than the CPU, what you have can matter. Video decode only matters if you are doing a lot of it. Some of the dedicated cards can decode a lot of streams at once, whereas the integrated cards usually do less.

But if all it is doing is playing back a single video, so which you are matching audio, then integrated vs discrete should make no difference.

2 Likes

Thanks, guys! That helped a lot :+1:

The key to a good reproduction is the codec used and size of frame. According to your message you have a PC computer and I have a Mac. If I have to advise a Mac user I would tell him to use videos encoded in ProRes proxy: mac processors are very optimised for this codec. I think that sure that there are goods and optimal codecs for PC . I’ve got a decklink monitor and Nuendo works better by playing the video directly

I think another question would be: are you planning to do any video editing on your system or is it going to be a strictly Nuendo station? If the first is true then you definitely need a dedicated GPU especially if you are doing any color correction, compositing, or transcoding. Please note a Decklink or an AJA card is primarily designed for video and audio playback on a dedicated monitor over SDI or HDMI. Lot of people confuse that as a GPU. CUDA and OpenCL video processing/rendering is handled via GPU (Nvidia, AMD).
That being said if you are getting a lot of film/tv work even if you are not editing the video, a dedicated video monitor and a playback card (like Blackmagic Design) is a must.
Nuendo supports the Blackmagic Design cards (Decklink, Intensity, Ultrastudio) .

Thanks! I do video editing and was already planning a GPU. This is my third PC upgrade at this point. I was just mainly wondering how Nuendo worked with dedicated GPU’s vs integrated and if that would influence the GPU decision at all. :+1:

If you do get a dedicated GPU for pro work, I’d get an nVidia. Reason is that so much stuff is CUDA only and even the things that are OpenCL often are OpenCL 3 which AMD doesn’t support yet. AMD has just kinda dropped the ball on general purpose compute type support so almost everything really like nVidia.

Of note is that Spectralayers 10 (with a lite version included in Nuendo, I believe) uses the GPU making those functions quite speedy compared to CPU.

2 Likes

That right there is one of those hidden gems I was expecting :joy:

Thanks, Steve! That’s great to know.