Computer simulations at the atomistic level are often called realistic as they offer, in principle, the possibility of reproducing in full the properties of a molecular system. Unfortunately, the number of atomic centers for molecules forming liquid-crystalline phases is normally so large, and correspondingly the number of molecules considered so low, that the proof of true realism, for example, the approximate reproduction of transition temperatures and of the relevant observables (such as order parameters together with their temperature dependence) has to our knowledge never been given to date. This is understandable in view of the complexity of liquid crystal molecules and probably depends on at least three issues: interaction potentials, sample size, and equilibration time. The force fields available have not been developed to reproduce liquid crystal properties, and only recently have some recommendations on the most appropriate choice amongst the many force fields available [1] been made. Sample sizes have varied from a number of particles N 50, [2±4] to 64 N 75, [5±10] to 100 N 125, [8, 11±15] and to N 200 molecules. [16] In a few recent cases, larger samples with N % 10 3 [17,18] particles have been simulated but without really studying the full temperature dependence and identifying transitions.At atomistic resolution, the observation time window in molecular dynamics (MD) calculations has also been quite short, ranging from less than % 600 ps [2, 3, 5±7, 10, 11, 14±18] to 1 or 2 ns. [4,9,13,15,16] Recently, a few simulations lasting up to 8 and 12 ns have appeared, [8,12] but these have involved the study of one, or very few, state points: again not sufficient to assess the ability of the methodology to predict transitions. The rather short runs are probably also connected to the contradictory preliminary findings on size dependence: thus in a recent study comparing N 118 and N 994 simulations, no size dependence was found: [18] in contrast to what was found on going from N 125 to 1000. [17] These strong limitations have been paralleled by the development of molecular-level models, where molecules are considered as simple objects, either just hard or endowed with attraction and repulsion, as in the Gay ± Berne models.[19] These approximate models have proven extremely useful in studying the general properties of liquid crystals (LC), such as phase transitions and correlations, and in discussing the existence of new and yet undiscovered phases such as the biaxial [20] and the ferroelectric nematic, [21] suggesting design hints to the synthetic chemist.However, there is a class of problems and observations that are very important for the understanding of anisotropic fluids, and completely depends on an atomistic level description. The foremost example is probably the prediction of the nematic ± isotropic (NI) transition temperature T NI , a crucial element in the design of viable LC display materials that have to exist and operate in a certain temperature range, that is now tackled empirically. [22,23...