This document reports results of the core experiment tests CE3-1.2.1 and CE3-1.2.2, both treating affine linear weighted intra prediction modes. These modes generate the luma intra prediction signal out of one line of reference samples left and above a current block by a matrix vector multiplication and the addition of an offset. In the first CE, at most twelve multiplications per sample are needed and the total memory requirement is 0.273 Megabyte of memory. The reported results are -1.36% luma BD-rate-gain for the AI and -0.85% luma BD-rate-gain for the RA configuration. The encoding time overheads are 153% resp. 113%. In the second CE, at most four multiplications per sample are needed and the total memory requirement is 0.018 Megabyte of memory. The reported results of the second CE are -0.95% luma BD-rate-gain for the AI and -0.57% luma BD-rate-gain for the RA configuration. The encoding time overheads are 153% resp. 110%. It is reported that in terms of operational complexity, the proposed intra prediction modes of the second CE do not exceed the conventional intra prediction modes and that in terms of memory requirement they do not exceed the CPR tool when restricted to the current CTU. For the second CE, two further results with reduced encoder complexity are presented: -0.85% and -0.57% luma BD-rate gain at 136% resp. 110% encoder overhead for the AI resp. RA configuration and -0.75% resp. -0.49% luma BD-rate gain at 127% resp. 106% encoder overhead for the AI resp. RA configuration.