For running on AMD CPU, ifort requires use of a compatible -m option (or no architecture option). The -x options won't work, and the -ax options will take the secondary code path (sse2 by default).
Latest AMD CPU versions should work with a variety of architecture options, including -mavx. Any AMD CPU produced in recent years should work with -msse3 (or the default -msse2).
You may require -imf-arch-consistency=true in case the default automatic run-time selection of math library is unsatisfactory. This option has been improved in recent ifort releases.
You can take out a 30-day trial license to see if it works for you.
What Tim says is correct, though the level of detail may have confused you. If you don't use any of the options for targeting specific processor kinds, then you'll have no problem. If you use -xHost to say "optimize for the processor I'm compiling on", you'll have no problem. You can also use the -m options to target specific instruction sets without requiring an Intel CPU.
If you look at the performance comparisons at polyhedron,com you'll see that they got the best performance on AMD processors using the Intel compiler.
