Accurate Value at Risk estimations of investment portfolio is crucial in the market risk management context. In recent years, traditional approaches to the VaR estimation, as variance-covariance method, are complemented by more flexible approaches based on the application of copulas. The aim of this paper is to compare the application of EVT-copula approach using five different copula functions with the traditional variance-covariance method. The results show a lack of precision VaR estimates obtained by using the variance-covariance method, particularly in high volatility periods of returns. Conversely, the methods based on the extreme value theory provide more accurate outputs in these periods.