There is a number of testing Lack-of-Fit procedures available in the literature, each being proposed to deal with a particular situation. While these existing tests might do a reasonable job, their disadvantages and drawbacks have also been acknowledged. In this paper, we propose an exact F-test that applies to any situation even if replicates or near replicates are not present. Its implementation is easy, and found to be extremely powerful compared with other tests. Its use is also developed when multidimensional covariates occur.