By implementing a large-area, gain-stabilized microcalorimeter array on an electron beam ion trap, the electron-impact excitation cross sections for the dominant x-ray lines in the Fe XVII spectrum have been measured as a function of electron energy establishing a benchmark for atomic calculations. The results show that the calculations consistently predict the cross section of the resonance line to be significantly larger than measured. The lower cross section accounts for several problems found when modeling solar and astrophysical Fe XVII spectra.