The study investigated the joint contribution of the self-regulated learning (SRL) and individual differences approaches to the prediction of university students' grade point average (GPA) obtained at three separate time points throughout their degree (3 years). We assessed cognitive (i.e., previous academic performance, cognitive ability, and cognitive SRL strategies) and non-cognitive variables (i.e., personality, trait emotional intelligence, motivation, and non-cognitive SRL strategies) in a sample of Spanish undergraduates. Results showed that GPA correlated with previous academic performance (i.e., combination of high school's GPA and college admission test score), academic self-efficacy, academic engagement, SRL strategies, and conscientiousness. Hierarchical regression analyses indicated that non-cognitive factors (i.e., academic engagement, academic self-efficacy, regulation of behavior and context, and conscientiousness) alone explained 17-25% of the variance in GPA across three years, and previous academic performance accounted up to an additional 25% of the variance, jointly reaching an explained variance of up to 50% in GPA. Specifically, academic engagement and regulation of behavior and context demonstrated incremental validity over and above cognitive predictors such as previous academic performance, inductive reasoning and regulation of cognition and metacognition. The role of intelligence, whether cognitive or emotional, was not as obvious as a predictor. Two nested structural equation models explained about 27-29% of the variance in a latent GPA factor exclusively from a proxy of a global variable of non-cognitive factors as a latent predictor, which is a novel and promising proof of its robust criterion validity. Implications and recommendations for future studies are discussed.