Monte Carlo simulations are regarded as the most accurate method of solving complex problems of radiation transport. Therefore, they have great potential to realize more exact dose calculations for treatment planning in radiation therapy. However, there is a lack of information on how correct the results of Monte Carlo calculations are on an absolute basis. A practical verification of the calculations can be performed by direct comparison with a benchmark experiment. Thereby, the uncertainties of the experimental result and of the simulation also have to be considered to make a meaningful comparison between the experiment and the simulation possible. This dissertation presents a benchmark experiment and its results, including the uncertainty, which can be used to test the accuracy of Monte Carlo calculations in the field of radiation therapy. The experiment was planned to have parallels to clinical radiation therapy, among other things, with respect to the radiation applied, the materials used and the manner of dose detection. The benchmark experiment aimed at an absolute comparison with a simulation result and because of this it was necessary to use a special research accelerator as a radiation source in the experiment. The accurate characterization of the accelerator beam was a precondition to define a realistic radiation source for the Monte Carlo simulation. Therefore, this work also deals with the characterization of the source and investigations regarding the X-ray target used. Additionally, the dissertation contains the verification of the widely used Monte Carlo program EGSnrc by the benchmark experiment. The simulation of the experiment by EGSnrc, the results and the estimation of the uncertainty related to the simulation are documented in this work.The results and findings of this dissertation end in a comparison between the results of the benchmark experiment and the corresponding calculations with EGSnrc. The benchmark experiment and the simulations were carried out for two different types of ionization chambers. The simulation results on average lie, for both cases, within the range which is set by the experimental result and its uncertainty < 0.7 %. The uncertainty of the EGSnrc calculation can be specified as < 1.0 %.