So, I've had this question floating around in my head for a while. Given the high amount of power that generative AI uses, why aren't we using ARM based chips for data centers? They consume far less energy than x86 based chips and the technology has improved exponetially in the past ~6yrs. If it's because of hardwar constraints, what exactly would need to change for ARM chips to be considered a viable chip architecture for this purpose?
[link] [comments]