Practical limit on matrix size for singular value decomposition (SVD) in R or MATLAB?

Hi folks,

I'm hoping people can give me a ballpark idea of the upper size limits on matrices for SVD in R or MATLAB. I am a newbie to using SVD in analyses, and I don't have a good intuition for what size input data is reasonable for this type of analysis. I have some imaging data--on the order of 100 images by 100,000 elements per image--which I would like to perform SVD or some other type of dimensionality reduction on. Is this reasonable to run on a decent newish Linux or Mac desktop? (Let's say 8 GB of RAM, 64-bit OS.) Execution speed is not so important to me, although for my purposes, several hours or more would be off-putting.