\indent many data-related applications utilize principal component analysis and/or data dimension reduction techniques that require efficiently computing dominant part of singular value decompositions (svd) of very large matrices which are also very dense. in our talk, we introduce a limited memory block krylov subspace optimization method which remarkablely accelerate the traditional simultaneous iteration scheme. we present extensive numerical results comparing the algorithm with some state-of-the-art svd solvers. our tests indicate that the proposed method can provide better performance over a range of dense problem classes under the matlab environment. we also present some convergence properties of our algorithm.