Inductors realized with high permeable MnZn ferrite require, unlike iron-powder cores with an inherent distributed gap, a discrete air gap in the magnetic circuit to prevent saturation of the core material and/or tune the inductance value. This large discrete gap can be divided into several partial gaps in order to reduce the air gap stray field and consequently the proximity losses in the winding. The multi-gap core, realized by stacking several thin ferrite plates and inserting a non-magnetic spacer material between the plates, however, exhibits a substantial increase in core losses which cannot be explained from the intrinsic properties of the ferrite. In this paper, a comprehensive overview of the scientific literature regarding machining induced core losses in ferrite, dating back to the early 1970s, is provided which suggests that the observed excess core losses could be attributed to a deterioration of ferrite properties in the surface layer of the plates caused by mechanical stress exerted during machining. However, in a first experimental analysis no structural evidence for a deteriorated layer close to the surface is identified by means of Scanning Electron Microscopy (SEM). Therefore, in a next step, a new calorimetric measurement setup based on temperature rise monitoring is proposed in this paper in order to quantify and differentiate between core losses associated with the bulk and the surface of the ferrite plates and therefore to pinpoint the measured excess core loss to shallow layers of ferrite with deteriorated magnetic performance. Electrical measurement of the surface related core losses utilizing the widely accepted twowinding wattmeter method with reactive power compensation is outlined in the appendix but was not employed in this work due to comparably low measurement accuracy. By means of the proposed measurement technique, the bulk and surface core loss density of the MnZn ferrite material 3F4 from FerroxCube was determined for sinusoidal flux density amplitude varying from 75 mT up to 200 mT and excitation frequencies ranging from 200 kHz to 1 MHz. The measured core loss densities (W/cm 3 ) show good agreement with the Steinmetz model provided by the manufacturer validating the proposed calorimetric core loss measurement technique. The measured surface loss density (W/cm 2 ) can also be well predicted with a Steinmetz model, whereby the frequency exponent α in the surface is slightly smaller and the flux density exponent β is slightly larger compared to the Steinmetz parameter of the bulk ferrite. It is shown that the ratio between surface and bulk core losses of a composite core assembled from individual plates is only a function of plate thickness and does not depend on the actual cross section