I recently ported some code to use AVX intrinsics. The performance gain is somewhere from 30x to 3000x depending on the size of the array. The larger the array the more performance gain I get. It’s linear algebra but not related to matrix multiplication that I could mention.