We present a detailed spectroscopic and timing analysis of X-ray observations of the bright radio-to-gamma-ray emitting pulsar PSR B0656+14, which were obtained simultaneously with eROSITA and XMM-Newton during the Calibration and Performance Verification phase of the Spektrum-Roentgen-Gamma mission (SRG). The analysis of the 100 ks-deep observation of eROSITA is supported by archival and simultaneous observations of the source, including XMM-Newton, NuSTAR, and NICER. Using XMM-Newton and NICER we firstly established an X-ray ephemeris for the time interval 2015 to 2020, which connects all X-ray observations in this period without cycle count alias and phase shifts. The mean eROSITA spectrum clearly reveals an absorption feature originating from the star at 570 eV with a Gaussian σ of about 70 eV, tentatively identified earlier in a long XMM-Newton observation (Arumugasamy et al. 2018). A second absorption feature, described here as an absorption edge, occurs at 260 − 265 eV. It could be of atmospheric or of instrumental origin. These absorption features are superposed on various emission components, phenomenologically described here as the sum of hot (120 eV) and cold (65 eV) blackbody components, both of photospheric origin, and a power-law with photon index Γ = 2 from the magnetosphere. We created energy-dependent lightcurves and phase-resolved spectra with high signal-to-noise. The phase-resolved spectroscopy reveals that the Gaussian absorption line at 570 eV is clearly present throughout ∼60% of the spin cycle, but undetected otherwise. Likewise, its parameters were found to be dependent on phase. The visibility of the line strength coincides in phase with the maximum flux of the hot blackbody. If the line originates from the stellar surface, it nevertheless likely originates from a different location than the hot polar cap. We also present three families of model atmospheres: a magnetised atmosphere, a condensed surface, and a mixed model, which were applied to the mean observed spectrum and whose continuum fit the observed data well. The atmosphere model, however, predicts too short distances. For the mixed model, the Gaussian absorption may be interpreted as proton cyclotron absorption in a field as high as 10 14 G, which is significantly higher than that derived from the moderate observed spin-down.