Your expectations may be unrealistic, or your code may have errors. Can you show us a test case? You usually won't get the exact same floating point results as these can vary depending on order of operations, extended precision and more. CVF always used "X87" instructions for floating point, often doing calculations in greater precsion than what you declared. Intel Fortran, by default, uses the faster SSE instructions, but always uses declared precision. You can force Intel Fortran to use X87 by adding the /arch:ia32 switch, but the program will run slower.
