This proposal evaluates the coding performance and runtime for the SAD-based early termination technique in VTM-5.0. The SAD-based early termination technique would turn off DMVR for a bi-prediction CU when the sum of absolute difference between the two prediction blocks of this CU falls below a predefined threshold value, that is, “4 * subblock width * ( subblock height >> 1)” or TH=4 in short. Several configurations of TH (i.e., 0, 8, 12, 16) other than the one (i.e., 4) used in the current VVC draft standard are tested, one at a time, to explore the trade-offs between coding gain and runtime impact. Experiments were conducted on top of VTM-5.0 with Random Access testing condition. Performance results are summarized as follows.