How come things like ROCm and the intel stack aren’t able to rapidly improve their software ecosystems to be a match for CUDA? Until the software from other vendors catches up with NVIDIA, they’re always going to get away with charging a massive premium on their “it just works” products. This is a genuine question, I’m using NVIDIA and Apple Silicon for my AI adventures thus far, but like everyone else on this subreddit, I want the prices to be more affordable. They won’t get that way until th...