Differentiation of brain signal variability across different cognitive states has been hypothesized to facilitate adaptation to changing task demands, but why signal variability should be higher or lower on a given task remains unknown. We hypothesized that the level of brain signal variability should mirror the feature density of sensory input, especially in high performers. To test these hypotheses, we had 46 healthy older adults passively view face and house stimuli during fMRI. We first used a computational model of the ventral visual stream (HMAX) to decode the feature content of all face and house images seen by participants; model results revealed that house images were much more feature-rich than faces, particularly for V1- and V2- like model layers. Using fMRI, we then found that participants whose V1/V2 brain signal variability increased the most in response to more feature-rich visual input (houses vs. faces) also exhibited faster and more stable behavioral performance on a comprehensive battery of offline visual tasks. We conclude that the ability to align visuocortical signal variability to the density of visual input may mark heightened trait-level behavioral performance in older adults.