A 4DVAR version of NCEP's data assimilation system, known as GSI, is gradually getting ready for testing in the environment of EMC computational framework and operationally used datasets. This was achieved through a collaborative effort of the research team gathered around development of GSI, most notable through a collaboration between EMC/NOAA and GMAO/NASA. While there are still unresolved issues, primary in the area of computational efficiency, this paper presents a series of preliminary data assimilation results involving a 4DVAR version of GSI that relies on a simple perturbation model used at NCEP for calculation of trajectory. These tests include a single observation test case, a series of standalone analyzes and a full cycling experiment. The main focus is on the comparison of 4DVAR against 3DVAR, in a data assimilation with both standard and various types of satellite observations.