Why deep learning support NIVIDEA rather than AMD what are the requirements that AMD doesn't fulfil ?

Answer 1

Might help
http://timdettmers.com/2018/11/05/which-gpu-for-deep-learning/
https://blog.slavv.com/picking-a-gpu-for-deep-learning-3d4795c273b9

Answer 2

I'm not sure that I'd say "deep learning" (as a movement, or entity) supports one company's tech over another, but one thing that comes to mind as beneficial in ML (etc) is how easy nVidia made it to access their GPU's (Graphics Processing Units) in a "general purpose" way via CUDA.

This was important not because it was intrinsically faster to access a GPU (or easy to program for a GPU like for CPUs) but because it opened up *massive parallelization* -- vs the paltry boost we get with, e.g., CPU hyperthreading. Put another way, GPUs are especially suited to applying "one instruction" to massive amounts of data, while CPUs are suited to applying "many instructions" to small amounts of data.

[Intended as a more general answer because I haven't looked at this recently; other members may have better info]