When diagnosing Rotator Cuff Tear(RCT), magnetic resonance imaging(MRI) scanned 3D data is largely used. Compare to 2D-based medical image, 3D data can offer more detail information and intuitive visualization of patient's condition. For example, three-dimensionally visualized MRI volume data of rotator cuff area can reveal not only presence or absence of the rupture, but also clear position and shape of it. However, only 2D-based slice images are used to diagnose RCT without taking advantage of 3D volume at the general medical field. Most of Convolutional Neural Network(CNN)-based medical image diagnosing methods are also using 2D data. We have proposed a RCT diagnosis method using 3D CNN that uses 3D information and take advantages of it to do the same task. The Voxception-Resnet(VRN) network was used to classify if volume has RCT or not. The data was preprocessed and resampled to 64x64x64 volume. and showed about 80 percent of accuracy when diagnosing test data. The 3D Class Activation Map(CAM) was also applied to visualize approximate location and shape of RCT. Our proposed method can automatically diagnose the presence of RCT using 3D data, and also visualize the activation map in 3D.This is less onerous and time-consuming than using 2D data.