I say we take a lesson out of AMD64 architecture and combine all microprocessors onto a single die...I'm sure this would in time be possible, have one core, or even more, dedicated GPUs, dedicated CPU cores, dedicated PPU cores...each core has a dedicated amount of cache size that would be large enough to replace memory
Basically just having one chip with possibly hundreds of dedicated cores communicating between each other at literally lightning data transfers speeds(see memory controller for AMD64s) and just have the core connected to external devices like hard drives and such...now THAT would be ridiculously fast
Basically just having one chip with possibly hundreds of dedicated cores communicating between each other at literally lightning data transfers speeds(see memory controller for AMD64s) and just have the core connected to external devices like hard drives and such...now THAT would be ridiculously fast