Saw this awesome little article on Extreme Tech. Engineers have managed to take an AMD “Trinity” APU and reduced the CPU portion to a pre-fetch/decode unit, while the GPU is used as the main computation unit. This brings massive gains in floating point performance as GPU’s are massively parallel processors.
These engineers out of North Carolina State University had their research Sponsored and Co-Authored by AMD. This goes to show how AMD is taking another step toward executing their new innovative HSA (Heterogeneous System Architecture) scheme. This involves a fully modular APU, that has parts (CPU,GPU etc) essentially picked and utilized depending on the use of the chip. This means we could be seeing more efficient use of the APU to make for some massive performance games. AMD has been encouraging this integration ever since the release of their first “Bobcat” based APU. Now all they need is to get many developers and programmers on board.
This is the pure innovation that I expected from AMD. They will break the linear CPU trend and create something that can be used for a wide range of mobile, desktop, and server uses. This puts pressure on Intel to develop a better graphics architecture to feature in their CPU’s in the future.