In VTM-3.0, the linear model for CCLM modes are derived by a max-min method. 128 neighbouring luma samples are down-sampled and 127 comparisons are performed on them to find the minimum and maximum values for a 64×64 chroma block. Besides, a lookup table with 2K bytes are stored on ROM to mimic the division operation. In this contribution, it is proposed that four neighbouring luma samples are down-sampled and three comparisons are performed on them for a 64×64 chroma block. And the lookup table size is reduced to be 128 bytes. Simulation results reportedly show 0.03% and 0.01% BD-rate loss under All Intra (AI) and Random Access (RA) configurations, respectively.