Intel's 14th gen CPU features dedicated VPU to offload AI tasks!

“Intel detailed some aspects of it’s 14th Gen Meteor Lake CPU’s AI smarts. All Meteor Lake SKUs will use a dedicated visual processing unit (VPU) that works together with the CPU and GPU and helps offload intense AI tasks”

Will TOPAZ be using this New AI Technology in the new 14th Gen Intel CPU’s to speed up the rendering time of 8K video upscaling in TOPAZ Video AI?

I was about to buy a new Intel I9-13900K CPU, but If the new 14th gen CPU will make rendering 8K video significantly faster, then I would like to wait and purchase an I9-14900K, instead.
I have a NVIDIA RTX 4090 video card, if that makes any difference.

Can you please tell me what I should do?
Thanks so much for your time, info and advice! :slightly_smiling_face:

I really :heart_on_fire: TOPAZ VIDEO AI
It is making my old videos look Gorgeous upscaled to 8K on my new Samsung 8K Tv!
Thanks

1 Like

I’m not sure how that’s going to work with Nvidia cards as directml does not support FP16 and only works with directx12.
I’m going to play around with ONNX Runtime on DirectML and TensorFlow with DirectML but if this tech lives up to what it claims it could pull me away from AMD for my next CPU upgrade.

Since the Intel 14th gen processors are not out yet, there’s no way for us (the average consumer) to know how well the VPU performs.

Everything below relies on some assumptions which may be incorrect:
Along with that, TVAI doesn’t appear to write their own model runner. Instead it seems they create their own AI models then run them on users computer using standard libraries for running AI models (E.G. ONNX, OpenVino, TensorRT, CoreML, etc).
As a result of this approach, it’s not really up to TVAI to decide on whether or not VPU support is added. It’s more up to the model runner developers (and Intel) to add support to the model runners, then for TVAI to update their version of the model runner to support the VPU.

Once all that is setup, the TVAI developers needs to then run tests to ensure the VPU/VPU backend provides results close enough to their “reference implementation”, verify there aren’t any bugs, then make a decision on if it’s worth enabling.

For example, AMD added AI accelerators to their RX 7000 series. To my knowledge TVAI still doesn’t use them. And I believe it’s because they discovered the performance of the AI accelerators wasn’t as good as they expected, got in contact with AMD, and AMD told them to hold off enabling the AI accelerators on RX 7000 series until AMD looked into the issue and identify the cause.

2 Likes

I would not wait because once you want to wait, you can wait forever because there will always come up something new. I personally grabbed the 13900k which is damn fast. Also I think Video AI is not yet using the RTX build in a. i. technologies (don’t proove me wrong) so I would not ask for a certain CPU type that is not yet available!

Great questions! I am having our development team look into this since I have no knowledge here since these CPUs are not yet released.

2 Likes

Awesome!
I really appreciate that! :heart_on_fire:
I look forward to hearing what you learn. :+1:t2:

ok, thanks :sunglasses:

So, basically, if they ever do use this feature it probably will not be anytime soon. LOL
Would be super cool if it does happen!
Thanks for your reply and info!
Take care

Thanks for telling about your experience!
Helps, alot! :slightly_smiling_face:

Intel is probably working on adding VPU support to at least one of the AI backends, so support in the AI backend will probably come a few weeks after Intel 14th gen releases (This is just speculation).

After that, the rest of the work is on Topaz Labs. And I don’t know their time frame for supporting things like this.

We work closely with Intel and I should be able to have an answer fairly soon :slight_smile:

1 Like

Yes, it will support the VPU when it is available :slight_smile:

1 Like

So shouldn’t Topaz also use the a. i. abilities included in the RTX series of GPU? :slight_smile:

We do not :frowning:

and the reason being?

We use the tensor cores in RTX GPUs. We do not plan to use Nvidia specific AI models like DLSS etc. We would ideally like to stick with our models and have the same models no matter the hardware or OS.

4 Likes

That is awesome! :partying_face: :balloon:
Thanks so much, ida, for responding to my question and finding an answer so quickly!
You are the best! :boom:

2 Likes

Hi suraj,
Thank you for sharing what you know on this subject!

I have 2 questions for you…

  1. When the 14th gen Intel CPU is released and Topaz Video AI is supporting the new A.I. VPU abilities to increase the rendering speed…
    will this still work if I am using my NVIDIA RTX 4090 GPU with the new 14th gen CPU, or will it only work if I am using an Intel GPU with it?

  2. If it currently takes me about 3 hours to render a 4 minute 8K video using the Proteus AI model…
    will the new Intel A.I. VPU technology speed up this rendering time at all?
    Say, like only 2 hours or less to render? :sunglasses:

Thanks for any input you can share with me!

I really love your software.
Maybe when I have completed a video, I can link my uploaded YouTube video for you to see how amazing and beautiful your Video AI upscaled my old music concert dvd’s to 8K! :boom:
I can help promote your software! LOL

maybe delay the questions to the point when the new generation of cpu exists. until then it is probably too much of a speculation, so lets concentrate on existing technologies to improve topaz products. :slight_smile:

1 Like

Exactly!