[TH] Windows 8: Does AMD's Bulldozer Architecture Benefit? - Page 4

Originally Posted by sumitlian 

Just because FX fails in Cinebench single thread benchmark doesn't mean that it would always perform lower. Cinebench depends on the FPU performance and a single core of FX can only utilize the half of 256 Bit FPU because of shared FPU unit between both cores in a Piledriver module. And when its about multithread FX wins because of it now uses all four 256 Bit FPU, and then because of its higher clock rate FX wins over i5.

AND as I've stated above that "Just because FX fails in Cinebench single thread benchmark doesn't mean that it would always perform lower in single thread" I believe is true because most real world single thread programs don't expect much from FPU. They can perform at its optimum speed by CPU's higher Integer Bandwidth. And most you should know FX's each core has two Integer pipeline while i5 has ONLY ONE. Clearly FX has massive integer bandwidth than i5 and some i7 and indeed FX wins in most Integer test benchmark. This proves FX may be even feel faster in most single thread apps (especially those which hasn't been designed to use Floating Point Unit)...<< This is what people don't understand.

And even if Windows explorer and Internet browsers are single thread, they can perform more quickly with FX because of FX's Dual Integer Pipeline per Core, obviously FX can handle more threads per core than an i5. They will feel equal or may be faster than i5, Also you can not forget Internet browsers can now use GPU for 2D/3D workload. AND I BELIEVE FX WILL REALLY FEEL FASTER.

Behind the FP scheduler are two 128-bit wide FMACs. AMD says that each thread dispatched to the core can take one of the 128-bit FMACs or, if one thread is purely integer, the other can use all of the FP execution resources to itself.