If I save the EOF of smaller domians, 0-10, 60S-50S (say) and later combine these files for the whole domain, would the result be the same as finding the EOF of the whole domian.
No, the combined EOF would not be the same as the EOF stemming from the whole sample. This would only work if the variability in each subdomain was independent of that in all the other subdomains.
If you find that ferret's performance in computing EOFs is poor, you might want to use, (a), a coarser time / space resolution (depending on what you're looking for) or, (b), a less automatic but more customizable setup (perhaps in Matlab) to only compute that part of the EOF / PC pairs you really need. I think, Dennis Hartmann's website http://www.atmos.washington.edu/~dennis/ contains lecture notes on Objective Analysis including a detailed discussion of the computation of EOF's in Matlab.
Willi