How can I speed up the eigen value and eigen vector computations for a non-sparse matrix?

I only need the eigen vectors corresponding to the 10 smallest eigen values. I do not want to use eigs because its results can vary. My matrix is 10000 x 10000 in size. I tried the parallel computing toolbox but get an error ""Bad version or endian-key" and somewhere under all the errors it also says Out-of-Memory. I am working with images, hence the size of the matrix.

Answer 1

[Andrew Knyazev]

check http://www.mathworks.com/matlabcentral/fileexchange/48-lobpcg-m