Convergence of Conjugate Gradient

• The number of conjugate gradient iterations needed for given accuracy for simple problems (positive definite matrix A) is given by

  

  where h is grid size

• Remember SOR was overrelaxation parameter w chosen correctly, which is not easy.

• Jacobi and Gauss Seidel have a number of iterations which is much more

• Conjugate gradient iterations are more complicated and time consuming than those for Jacobi or Gauss Seidel