NVidia got there early with their CUDA API.
That’s been around for decade(s), which enabled all sorts of crazy GPU usages beyond just graphics.
Due to that, NVidia held the datacenter/professional scene exclusively for a long time.
As a result, their professional cards and related drivers have been industry standard.
I have no doubt that AMD is better, but so much (non-mainstream) software is built against NVidia drivers, CUDA etc., that will be slow to change until the cost of implementing similar for AMD outweighs “just sticking with NVidia”.
The classic “Nobody ever got fired for buying IBM”
NVidia got there early with their CUDA API.
That’s been around for decade(s), which enabled all sorts of crazy GPU usages beyond just graphics.
Due to that, NVidia held the datacenter/professional scene exclusively for a long time.
As a result, their professional cards and related drivers have been industry standard.
I have no doubt that AMD is better, but so much (non-mainstream) software is built against NVidia drivers, CUDA etc., that will be slow to change until the cost of implementing similar for AMD outweighs “just sticking with NVidia”.
The classic “Nobody ever got fired for buying IBM”