After the discovery of GRS 1915+105 (Castro-Tirado et al. 1992) we obtained pointed ROSAT observations every six months (12 until now). The flux in the ROSAT (0.1–2.4 keV) band is strikingly different from the simultaneous BATSE (25–50 keV) flux which was obtained by integrating the best fit power law (Fig. 1). Motivated by the different intensity evolution in the soft and hard X-ray band we have selected BATSE monitoring data collected simultaneously to ROSATdata and performed joint spectral fitting with XSPEC. As a result, we never got an acceptable fit (see Fig. 2): The BATSE power law (upper dotted line) is too steep to match the ROSAT band, and even allowing for an increased absorbing column (lower dotted line) does not solve the problem. Alternatively, neither a thermal bremsstrahlung fit (solid line) nor a power law fit (lower dash-dot line) to the ROSAT data match the BATSE flux. The upper dash-dot line is a -2.5 powerlaw which would match the BATSE data while giving too much 1–2 keV emission. A similar, but less stringent result is obtained when folding the best fit BATSE power law models with the HRI detector response to compare the expected count rate with the observed one. We therefore conclude that the spectrum during all simultaneous ROSAT/BATSE observations seemingly consists of two different spectral components.