Starting from XMM-Newton EPIC-PN data, we present the X-ray variability characteristics of PKS 2155-304 using a simple analysis of the excess variance, \xs, and of the fractional rms variability amplitude, fvar. The scatter in \xs\ and \fvar, calculated using 500 s long segments of the light curves, is smaller than the scatter expected for red noise variability. This alone does not imply that the underlying process responsible for the variability of the source is stationary, since the real changes of the individual variance estimates are possibly smaller than the large scatters expected for a red noise process. In fact the averaged \xs and \fvar, reducing the fluctuations of the individual variances, chang e with time, indicating non-stationary variability. Moreover, both the averaged \sqxs (absolute rms variability amplitude) and \fvar show linear correlation with source flux but in an opposite sense: \sqxs correlates with flux, but \fvar anti-correlates with flux. These correlations suggest that the variability process of the source is strongly non-stationary as random scatters of variances should not yield any correlation. \fvar spectra were constructed to compare variability amplitudes in different energy bands. We found that the fractional rms variability amplitude of the source, when significant variability is observed, increases logarithmically with the photon energy, indicating significant spectral variability. The point-to-point variability amplitude may also track this trend, suggesting that the slopes of the power spectral density of the source are energy-independent. Using the normalized excess variance the black hole mass of \pks was estimated to be about 1.45×108M⨀. This is compared and contrasted with the estimates derived from measurements of the host galaxies.