We use the X-ray power spectrum of the soft X-ray background ROSAT all-sky survey in the R6 band (approximate to0.9-1.3 keV) to set an upper limit on the galaxy cluster power spectrum. The cluster power spectrum is modelled with a minimum number of robust assumptions regarding the structure of the clusters. The power spectrum of ROSAT sets an upper limit on the Omega(m)-sigma(8) plane that excludes all the models with sigma(8) above sigma(8) = 0.5 +/-(+0.08)(-0.05) Omega(m) -(0.38 +/- (0.01))(0.05) in a flat Lambda-cold dark matter universe. We discuss the possible sources of systematic errors in our conclusions, mainly dominated by the assumed L-x-T relation and some cluster signal subtracted during the point source removal. The uncertainties due to removed clusters can move our upper limit as much as 0.2 units upwards. The systematic bias introduced by uncertainties in the scaling relations is more difficult to quantify since these relations are not accurately known yet but we show how, alternatively, the L-x-T relation could be constrained using the X-ray power spectrum, if the cosmological model is known and provided a good estimate of the power spectrum of the diffuse X-ray background can be obtained from the data. Roughly accounting for all the systematics, our conclusions suggest that models with a low value of sigma(8) are favoured by the soft X-ray diffuse background. The approach presented in this work would be extremely powerful with high-quality data from Chandra and XMM with a better point source subtraction and going to smaller scales than those proved in this work.