Yea I don’t understand this chart at all
It’s as if a cuda core don’t evolve
4080 is drastically cut back, yet here we are, it’s trading blows with AMD’s flagship
In the end, the %cut from an architecture, doesn’t matter, it’ll depend on the foundry, node, yields, performance , etc. There’s such a thing as core stretching where you’re basically not gaining the performance linearly by having more cores. There’s a knee point and they optimize around that. A 4080 is not 53% of the 4090 performances.