CUDA Compute Power vs CUDA Cores for Video Enhance AI

Topaz Products Product Technical Support
1

I am trying to discern which attributes make a GPU work faster when upscaling video with VEAI.

Nvidia has two CUDA scores that I can find:

  1. CUDA Compute Capability
  2. CUDA Cores

My current GPU, the RTX 2060 Super, has the same Compute Capability as any card, even up to the RTX 2080 Ti, which is 7.5. (Comparison page here)

However it has many fewer CUDA cores. (Core comparison here)

So my simple question is this. If I buy a bigger, faster GPU that has more CUDA cores, but the same CUDA Compute Capability, will VEAI run faster or not?

1 Like
2

I can tell you that my RTX 2070 Super is not faster than my former RTX 2060 non Super in Gigapixel when both are clocked the same. That being said, GP does not make use of Cuda (GPU load is identified as “3D” not “Cuda”).

3

Interesting. I know that VEAI uses CUDA entirely, so I’m really fascinated by what their answer to this will be. The 2080 ti has about twice the number of CUDA cores (2176 vs 4352), but they are listed as having the same CUDA “Compute Capability”, which I find very confusing. I’d like to know that I’m going to get some kind of speed increase for this processing if I’m going to spend $1200 on it.

4

I think cuda is important. when I looked at my Win 10 Task Manager, Up to 98% of the CUDA is used.
when using v.e.ai
(and 3D mark is improtant when using gigapixel)

6

CUDA’s computing power seems to refer to the version of sdk. This means that it has nothing to do with the number of Cuda. link here