They already exist and have for a while, you can run e.g. Windows and do your daily computing using one, they just cost an arm and a leg.
>>25614185
Not happening because bandwidth and latency can be traded and most tasks care about latency more than bandwidth, graphics and some GPGPU tasks are the odd ones out. Unified memory isn't going to be on the high end until a memory technology that is better than current technology and can't trade bandwidth for latency comes out.