The author questions why software ecosystems for AMD's ROCm and Intel have failed to rapidly improve to match NVIDIA's CUDA. It is argued that until competing vendors' software catches up, NVIDIA will continue to charge a massive premium for its convenient products. The poster identifies as a user of both NVIDIA and Apple Silicon hardware for AI development. They express a desire for more affordable prices within the market. The argument suggests that price reductions will only occur when genuine competition exists. This perspective highlights the current dominance of CUDA in the AI hardware landscape.
Question on why ROCm and Intel stacks lag behind CUDA in software ecosystem maturity
from English