Over the past two decades, significant progress in the thermal metrology for thin films and wires has enabled new understanding of the thermal conductivity of nanostructures. However, a large variation in the measured thermal conductivity of similar nanostructured samples has been observed. In addition to potential differences from sample-to-sample, measurement uncertainty contributes to the observed variation in measured properties. Many now standard micro/nanoscale thermal measurement techniques require extensive calibration of the properties of the substrate and support structures and this calibration contributes to uncertainty. Within this work, we develop a simple, direct differential electrothermal measurement of thermal conductivity of micro/nanoscale sample films by extending conventional steady state electrothermal approaches. Specifically, we leverage a cross-beam measurement structure consisting of a suspended, composite heater beam (metal on silicon) with the sample structure (silicon) extending at a right angle from the center of the heater beam, in a configuration similar to the T-type measurements used for fibers and nanotubes. To accurately resolve the thermal conductivity of the sample, the steady-state Joule heating response of the cross-beam structure is measured. Then, the sample is detached from the heater beam with a Focused Ion Beam (FIB) tool enabling direct characterization of the composite heater beam thermal properties. The differential measurement of the structure before and after FIB cut enables direct extraction of the sample thermal conductivity. The effectiveness of this differential measurement technique is demonstrated by measuring thermal conductivity of a 200 nm silicon layer. Additionally, this new method enables investigation of the accuracy of conventional approaches for extracting sample thermal conductivity with the composite beam structure and conventional comparative approaches. The results highlight the benefits of the direct differential method for accurate measurements with minimal assumptions.