How long until we see four gpus on a single pcb? Why can't you just put two on the same slice?
The fastest single-core 90nm AMD cpu did 3.0 stock. Their fastest 90nm dual-core cpu did 2.8 stock. Is it reasonable to infer that two gpus could be put on the same slice and run nearly as fast as if they each had their own slice?
Or better yet why not just double all the pipes and shaders and interconnecting stuff. Wouldn't that lead to a die twice the size allowing for twice as much heat transfer while avioding the SLI performance tax (~15±)?
Is everything titrated by the marketing dept?
The fastest single-core 90nm AMD cpu did 3.0 stock. Their fastest 90nm dual-core cpu did 2.8 stock. Is it reasonable to infer that two gpus could be put on the same slice and run nearly as fast as if they each had their own slice?
Or better yet why not just double all the pipes and shaders and interconnecting stuff. Wouldn't that lead to a die twice the size allowing for twice as much heat transfer while avioding the SLI performance tax (~15±)?
Is everything titrated by the marketing dept?