YOu need understand both use GDDR5 as shared memory between the cpu and gpu , i dont think the choice to use high latency ram is made without the API in mind.. basically the cpu will just support the first instructions landing, and most of the GDDR5 will be used for the gpu with Virtual memory in mind ( basically both cores can land any instruction, codes in the memory who been taken directly for the gpu and vice versa ) thoses consles are highly multithreaded and will use parrallel work as never before. Both support HSA instructions, last DX11.1 multithreading features, OpenCL ( and the list of library is long like an harm ( including PhysX3 on software ) and this with a simple and efficient API / OS dedicated to only one thing, make work this hardware .




Reply With Quote
Bookmarks