This market seeks to predict whether AMD will introduce a hardware accelerator product that competes directly with NVIDIA's offerings in the AI space before the year 2028. As AI technology rapidly evolves, the demand for specialized hardware to efficiently train and run AI models has escalated. NVIDIA currently leads this market with its powerful GPUs and dedicated AI accelerators. This market will track AMD's progress and potential announcements related to AI hardware accelerators, focusing on their performance, market adoption, and technological innovation. Considerations include advancements in chip design, partnerships, market strategy, and the broader competitive landscape in AI hardware. Resolve this market affirmative if, before 2028, AMD releases a product that is widely recognized as competitive with NVIDIA’s GPUs for ML/AI purposes
Related questions
@osmarks It is still a little early to know if it will gain a competitive market share against NVIDIA H100
@ManuelAngelSuarezAlvarez so wait, are you looking for a product to be competitive in terms of technology, or market share? I agree the mi300x should satisfy the question as written, though it's understandable to wait until release and third party benchmarks etc
@Tomoffer market share, on the other hand, is down to all sorts of weirdness like "relationships" and "marketing"
@Tomoffer oh and there's a reason George Hotz is putting AMD in tiny boxes, that's a different card and a different possible YES justification
@ManuelAngelSuarezAlvarez How much market share would it need? Like @gpuk says in the other comment thread, my understanding is that in terms of raw flops AMD is competitive with Nvidia, but that no-one uses AMD chips for AI training or data centers because AMD doesn't have anything competitive with CUDA.
@MichaelWheatley fwiw my personal, relatively-casual-so-far experience is that ROCm with torch "just works" 🤷♂️