Early diagnosis of brain tumor is decisive for improving patient survival rates and results. Among the various diagnostic methods available, MRI-based imaging is widely used for detecting brain tumor due to its detailed imagining of brain structure. Researchers have explored numerous deep learning techniques to improve the accuracy of brain tumor detection. Despite the advancement, only few studies have successfully classified brain tumors into four specific categories: healthy, pituitary, gliomas, and meningiomas. Addressing this gap, our research focuses on developing an efficient method to classify brain tumor into these four categories with high accuracy and processing time. To achieve this, we designed two distinct convolutional neural networks, for two different MRI image resolutions. The first CNN was trained on MRI images with image size of 150 × 150 pixels, while the second CNN was trained on images with resolution 100 × 100 pixel. These models were developed to evaluate the impact of image resolution on classification performance. The proposed CNN demonstrated the impressive classification accuracies. The CNN trained using 100 × 100-pixel size MRI image achieved an accuracy of 95.13% while CNN trained on 150 150 image size achieved accuracy of 92.2%. These results indicate that the CNN designed for lower resolution image not only offers high classification accuracy but poetically reduces the computational time. Along with accuracy, precision, recall, and F1 score are also validated using the proposed technique.