I am getting some tremendous performance out of the cores(with AVX)and the HD 3000 on the i5-2500k. I can understand why the cores with AVX are producing the performance , but can not explain why the HD 3000 is so quick.
Do the 12 Execution Units on the HD 3000 use an AVX-like structure to achieve their performance? I am comparing the Execution Unit(EU)s to Nvidia's Streaming Multiprocessor design with each EU containing a 256bit register set(or longer)to achieve this large computing throughput.
Where could I find out more about the architecture of these EU?
