Sounds strange. -fast is nothing else than a collection of the following options: -ipo O3 -no-prec-div -static -xHost. -xHost should identify the processor and set the right SSE switch (-xAVX for i7 2xxx).
-fast is a prototype switch that shall show best possible performance just "out of the box". On real projects you shall consider using proper switches that achieve best performance.
What happens if you use the full "fast" options list: -ipo O3 -no-prec-div -static -xHost? You may consider to use only a suitable subset of these options such as -O3 -xHost.