glibc authors recommend to avoid using i586-specific optimization for i686+ since these processors are quite different in architecture from pure i586 -- e.g. three pipelines instead of two perform better with one command stream instead of two: http://git.altlinux.org/people/ldv/packages/?p=glibc.git;a=blob;f=sysdeps/i386/i686/Implies Is it reasonable to "downgrade" optimization aiming at higher performance of legacy/lightweight systems?
tracked at https://bugs.launchpad.net/rpm/+bug/911012