We report a numerical test of the Adam-Gibbs relation for the TIP4P/2005 model of water. The configurational entropy is here evaluated as the logarithm of the number of different basins in the potential energy landscape sampled in equilibrium conditions. Despite the non-monotonic behaviour which characterise the density dependence of the diffusion coefficient, the Adam-Gibbs relation is satisfied within the numerical precision in a wide range of densities and temperatures. We also show that expressions based on the excess entropy (the logarithm of the number of sampled microstates in phase space) fail in the region of densities where a tetrahedral hydrogen bond network develops.