Beyond eliminating the critical slowing down, multigrid algorithms can
also eliminate the need to produce many independent fine-grid configu
rations for averaging out their statistical deviations, by averaging o
ver the many samples produced in coarse grids during the multigrid cyc
le. Thermodynamic limits can be calculated to accuracy epsilon in just
O(epsilon-2) computer operations. Examples described in detail and wi
th results of numerical tests are the calculation of the susceptibilit
y, the sigma-susceptibility, and the average energy in Gaussian models
, and also the determination of the susceptibility and the critical te
mperature in a two-dimensional Ising spin model. Extension to more adv
anced models is outlined.