We estimate the sensitivity of future galaxy surveys to cosmological parameters, using the redshift dependent angular power spectra of galaxy number counts, Cl(z1,z2), calculated with all relativistic corrections at first order in perturbation theory. We pay special attention to the redshift dependence of the non-linearity scale and present Fisher matrix forecasts for Euclid-like and DES-like galaxy surveys. We show that for surveys with photometric redshifts the analysis with the redshift dependent angular power spectra performs significantly better than the usual analysis based on the three-dimensional matter power spectrum P(k,z). For spectroscopic redshifts, however, the large number of redshift bins needed to fully profit from the redshift information, leads to a significant degradation of the Cl(z1,z2) analysis due to shot noise. We also identify surveys which can measure the lensing contribution and we study the monopole, C0(z1,z2).