Before the introduction of HAART, HIV-associated neurocognitive impairment (NCI) was recognized as an independent risk factor for death. Since 1996, we conducted a prospective study to assess whether NCI still represents a negative prognostic factor for mortality. Patients were administered measures of neurocognitive function (a battery of 17 neuropsychological tests), clinical and neurological evaluation, laboratory testing, and brain imaging studies. Among the 412 enrolled patients, 224 (54.4%) were neurocognitively impaired and 188 (45.6%) were neurocognitively unimpaired. A durable virological suppression under highly active antiretroviral therapy (HAART) was achieved by 63.3% of unimpaired patients and by 49.6% of impaired patients (p = 0.007). Overall, 47 deaths were recorded, 38 among impaired and 9 among unimpaired patients. At 84 months, the estimated survival proportions in impaired and unimpaired patients were 68.5% and 84.9%, respectively (p < 0.001). At univariate analysis the virological response to HAART was the variable most strongly associated with survival, since patients with virological failure had a nearly 10-fold increased risk of death than those with durable virological suppression (HR = 9.9, 95% CI: 3.9-25.0). After stratification for virological response to HAART, an increased risk of death for neurocognitively impaired patients was seen only among the 182 patients with virological failure (HR: 2.9, 95% CI: 1.2-7.1), while the survival probability of the 230 patients with durable virological suppression was not affected by neurocognitive impairment (p = 0.89). Our results highlight the clinical relevance of HIV-related central nervous system (CNS) involvement in the HAART era, and raise concerns regarding the clinical relevance of CNS involvement as potent antiretroviral therapies become less effective.