with the way things are going I think in a generation or two, GPUs will generate frames instead of raster render them. ie. you will buy a compute chip from nvidia and then the game devs will have a specifc model that has been trained to display a game. this will then be generated so fast it will be imperceptible from a frame rate and super smooth with infinite resolution and zooming if needed, with more detail being added the closer the zoom. current compute chips are still primitive, give it a decade or so and they'll precisely generate images without flaws realtime at a 120fps at 8K. it only looks flawed now because we are still learning how to improve and prompt it into getting it to show us exactly what we want to see this will massively drop the power usage and performance cost of displaying a game. stable diffusion with precise prompts is a far more efficient way to do game design. a custom game engine will then be made to precisely and accurately prompt a model into creating what is required. think unreal engine but with ai prompts to get it to do what you want. when this happens all these render based gpus will all seem outdated. imagine how unique games can become when AI can generate new scenarios based upon input and display them realtime in fluid motion with near instant latency interaction from controls/input (ie. making a character move or looking left/right/up/down). When this tech matures, we can expect hyperrealism gaming with higher resolution than real life.