GIORIE-SLAM is a globally optimized RGB-only dense SLAM system that uses a flexible neural point cloud scene representation to adapt to keyframe poses and depth updates without costly backpropagation. The system addresses the lack of geometric priors in RGB-only SLAM by integrating a monocular depth estimator and a novel Disparity, Scale, and Pose Optimization (DSPO) layer, which optimizes the pose and depth of keyframes along with the scale of the monocular depth. This approach improves reconstruction completeness and accuracy. GIORIE-SLAM benefits from loop closure and online global bundle adjustment, performing competitively or better than existing dense neural RGB SLAM methods in tracking, mapping, and rendering accuracy on datasets such as Replica, TUM-RGBD, and ScanNet. The method achieves lower trajectory errors and higher rendering accuracy compared to competitive approaches like GO-SLAM. The source code is available at https://github.com/zhangganlin/GIORIE-SLAM.GIORIE-SLAM is a globally optimized RGB-only dense SLAM system that uses a flexible neural point cloud scene representation to adapt to keyframe poses and depth updates without costly backpropagation. The system addresses the lack of geometric priors in RGB-only SLAM by integrating a monocular depth estimator and a novel Disparity, Scale, and Pose Optimization (DSPO) layer, which optimizes the pose and depth of keyframes along with the scale of the monocular depth. This approach improves reconstruction completeness and accuracy. GIORIE-SLAM benefits from loop closure and online global bundle adjustment, performing competitively or better than existing dense neural RGB SLAM methods in tracking, mapping, and rendering accuracy on datasets such as Replica, TUM-RGBD, and ScanNet. The method achieves lower trajectory errors and higher rendering accuracy compared to competitive approaches like GO-SLAM. The source code is available at https://github.com/zhangganlin/GIORIE-SLAM.