This contribution presents a super-resolution network that combines CNN with existing RPR functionality in VVC, called MMSDANet. In MMSDANet, we propose a new basic block, multi-mixed scale and depth information with attention block (MMSDAB) to extract multi-scale and convolutional layer depth information, and shared convolution is used to reduce network parameters. Compared with VTM-11.0 RPR anchor, MMSDANet achieves {-6.72%, -26.89%, -28.19%} and {-8.16%, -25.32%, -26.30%} BD-rate gains on average for {Y, Cb, Cr}, under RA and AI configurations, respectively. Compared with VTM-11.0 NNVC-1.0 anchor, MMSDANet achieves {-4.21%, 4.53%, -9.55%} and {-8.5%, 18.78%, -12.61%} BD-rate gains on average for {Y, Cb, Cr}, under RA and AI configurations, respectively.