An interesting probe of the nature of dark energy is the measure of its sound speed , c _ { s } . We review the significance for constraining sound speed models of dark energy using large neutral hydrogen ( HI ) surveys with the Square Kilometre Array ( SKA ) . Our analysis considers the effect on the sound speed measurement that arises from the covariance of c _ { s } with the dark energy density , \Omega _ { de } , and a time-varying equation of state , w ( a ) = w _ { 0 } + ( 1 - a ) w _ { a } . We find that the approximate degeneracy between dark energy parameters that arises in power spectrum observations is lifted through redshift tomography of the HI-galaxy angular power spectrum , resulting in sound speed constraints that are not severely degraded . The cross-correlation of the galaxy and the integrated Sachs-Wolfe ( ISW ) effect spectra contributes approximately 10 percent of the information that is needed to distinguish variations in the dark energy parameters , and most of the discriminating signal comes from the galaxy auto-correlation spectrum . We also find that the sound speed constraints are weakly sensitive to the HI bias model . These constraints do not improve substantially for a significantly deeper HI survey since most of the clustering sensitivity to sound speed variations arises from z \lower 3.225 pt \hbox { $ \sim$ } \hbox to 0.0 pt { \raise 1.161 pt \hbox { $ < $ } } 1.5 . A detection of models with sound speeds close to zero , c _ { s } \lower 3.225 pt \hbox { $ \sim$ } \hbox to 0.0 pt { \raise 1.161 pt \hbox { $ < $ } } 0. % 01 , is possible for dark energy models with w \lower 3.225 pt \hbox { $ \sim$ } \hbox to 0.0 pt { \raise 1.161 pt \hbox { $ > $ } } -0.9 .