We perform a set of non--radiative cosmological simulations of a preheated intracluster medium in which the entropy of the gas was uniformly boosted at high redshift. The results of these simulations are used first to test the current analytic techniques of preheating via entropy input in the smooth accretion limit. When the unmodified profile is taken directly from simulations, we find that this model is in excellent agreement with the results of our simulations. This suggests that preheated efficiently smoothes the accreted gas, and therefore a shift in the unmodified profile is a good approximation even with a realistic accretion history. When we examine the simulation results in detail, we do not find strong evidence for entropy amplification, at least for the high-redshift preheating model adopted here. In the second section of the paper, we compare the results of the preheating simulations to recent observations. We show -- in agreement with previous work -- that for a reasonable amount of preheating, a satisfactory match can be found to the mass-temperature and luminosity-temperature relations. However -- as noted by previous authors -- we find that the entropy profiles of the simulated groups are much too flat compared to observations. In particular, while rich clusters converge on the adiabatic self--similar scaling at large radius, no single value of the entropy input during preheating can simultaneously reproduce both the core and outer entropy levels. As a result, we confirm that the simple preheating scenario for galaxy cluster formation, in which entropy is injected universally at high redshift, is inconsistent with observations.