Its probably based on the SSE, the info I got wasnt that specific, only stating 'floating point'. although most developers that have a need for processing large arrays of data are now going over to SSE, so the slower performance on the amd would impact its value for those calculations. How much of an impact it woudl have would depend on the actual calculations themselves, so for short vectors, it may not impact the throughput that much, btu for large vectors, where the OOE really pays off for tight sequences of code, it woudl have a larger impact I would think.
