The VRAM of present GPU can be replaced by PCMem, which limits the change of protocol and interface between GPU chip and PCMem, which is the simplest and easiest way to apply PCMem from present computer architecture. <\/p>\n\n\n\n
With this GPU dedicated PCMem, the whole computer architecture can keep exactly same or even utilize present CPU, motherboard and Dram directly except redesiging GPU with dedicated PCMem.<\/p>\n\n\n\n
With GPU dedicated PCMem, the connection in GPU is MPUs to PCMem chip bus to GPU PCB bus, and GPU connects to motherboard in traditional or even present ways. <\/p>\n\n\n\n
Although this GPU dedicated PCMem make PCMem seem like for GPU only but PCMem actually is a revolutionary paradigm shift and breakthrough of Memory Parallel Computing from traditional Von Neumann Model, which Gemini agrees with.<\/p>\n\n\n\n
For the farther future, CPU and GPU can be integrated in same package to share same PCMem.<\/p>\n\n\n\n
Gemini’s comment on this:<\/p>\n\n\n\n
This is a very sharp, pragmatic commercialization strategy.<\/p>\n\n\n\n
It turns a revolutionary architecture into an immediately deployable product.<\/p>\n","protected":false},"excerpt":{"rendered":"
The VRAM of present GPU can be replaced by PCMem, which limits the change of protocol and interface between GPU chip and PCMem, which is…<\/p>\n