Well, most of the money and marketshare is sub-£250, high end to most non-enthusiast people is a £300 GPU, Raja knows this better than anyone since this is where GCN was always targeted, Tahiti and Polaris were arguably GCN at its best.
Intel would already have their work cut out for them scaling Gen11 up that far, and we already know the dGPU parts are a kind of jack-of-all-trades arch like Vega, I think it will be positioned as a kind of prosumer card personally.
Even NVidia are taking the approach of big not-so-gaming optimised compute heavy dies now with Turing. Pascal's approach was great for (price or die size)/perf but only for what we currently/used to consider traditional graphics pipelines, leaving more of the optimisation to the software side of things like with GCN/Turing is exactly what Intel tried to do with Larrabee(And Itanium) and it's been their approach with CPUs for decades but is only really viable recently with more diverse and optimised and increasingly complex and organic software stacks.
This is why Larrabee was technically the first DX12 card.
Intel would already have their work cut out for them scaling Gen11 up that far, and we already know the dGPU parts are a kind of jack-of-all-trades arch like Vega, I think it will be positioned as a kind of prosumer card personally.
Even NVidia are taking the approach of big not-so-gaming optimised compute heavy dies now with Turing. Pascal's approach was great for (price or die size)/perf but only for what we currently/used to consider traditional graphics pipelines, leaving more of the optimisation to the software side of things like with GCN/Turing is exactly what Intel tried to do with Larrabee(And Itanium) and it's been their approach with CPUs for decades but is only really viable recently with more diverse and optimised and increasingly complex and organic software stacks.
This is why Larrabee was technically the first DX12 card.
Last edited: