On 03/09/12 12:32 AM, Kaushal Shriyan wrote: > Not sure i understand about your earlier comment regarding pure floating > point compute, help me understand with some examples. mostly, numerical scientific processing, and various sorts of multimedia work, such as batch converting HDTV video formats, 3D animation rendering, etc. most architectures, there's only one FPU (floating point unit) per core... hyperthreading causes each core to act like two cores, which works great for integer, character, network, etc sorts of processing, but if the two threads are both trying to do steady floating point or MMX/SSE style processing, they have to share the same hardware and run slower. -- john r pierce N 37, W 122 santa cruz ca mid-left coast