Gradient algorithm is very important for the Hartmann-Shack sensor. The Hartmann-Shack sensor based on the normalized cross correlation (NCC) gradient algorithm shows strong noise immunity. In addition, it can detect both point source and extended target. A multi- core CPU (central processing unit) is used to implement NCC gradient algorithm to get better portability compared to the existing processor. The NCC algorithm is optimized and programmed. The program is improved by multi-core parallelization and advanced vector extension. When the template resolution is 9 pixel×9 pixel, the optimized program running on a 4-core computer, Intel (TM) Core (R) i7-3770k, spends about 340 μs to calculate the slope of 400 sub-apertures with 15 pixel×15 pixel. An adaptive optics system using optimized gradient program is built and experiences closed-loop test. The experimental results show that the CPU- based Hartmann- Shack adaptive optics system using NCC gradient algorithm can correct wavefront distortions effectively, even in strong noise environment.