This paper presents a method for prediction of the unequal admission performance of a double entry turbine based on the full admission turbine maps and a minimal number of unequal admission points. The double entry turbine has two separate inlet ports which feed a single turbine wheel: this arrangement can be beneficial in a turbocharger application; however the additional entry does add complexity in producing a complete turbine map which includes unequal admission behavior. When a double entry turbine is operated under full admission conditions, with both entries feeding the turbine equally, this will act effectively as a single entry device and the turbine performance can be represented by a standard turbine map. In reality a multiple entry turbine will spend the majority of time operating under varying degrees of unequal admission, with each entry feeding the turbine different amounts; the extent of this inequality can have a considerable impact on turbine performance. In order to produce a full map which extends from full admission through to the partial admission case (where one inlet has no flow) a large number of unequal admission data points are required. The paper starts by discussing previous attempts to describe the partial and unequal admission performance of a double entry turbine. The full unequal admission performance is then presented for a nozzled, double entry turbine. The impact of unequal admission on turbine performance is demonstrated. Under some conditions of operation, the turbine efficiency may be less than half that of the equivalent full admission case based on the average turbine velocity ratio. A method of using the steady, equal admission maps, with a limited number of unequal admission data points, to predict the full unequal admission behavior is presented. A good agreement is found when the map extension method is validated against the full unequal admission turbine performance measured on a test stand. In the prediction of efficiency a mean error of approximately 0.39% is found between the test stand data and the proposed extrapolation method, with a standard deviation of 2.79%. A better agreement is generally found at conditions of higher power.