Probably some confusion... but my understanding is the current P4 and Opteron can do two floating point operations per cycle... and the fused macro-ops in Conroe increase these to four, iff they are two multiplies followed by accumulates of the multiply results... effectively doubling the linpack result that can be achieved. If you have a different understanding of what Intel means by "fused macro ops" I would love to hear it. --Alan
Market Data 
Markets 
AI
