In addition to the other monks' advice, please take into consideration that hashes are not designed to be "conservative" with regard to memory consumption. They are designed to be fast. Perhaps the exception to this, is a relatively new optimization where the text for common hash keys is stored once IIRC.

A simple algorythm would be something along the lines of this untested, pseudo-code example:

my @uniques = ();

my $md5;

while (my $string = <FILE>) {
  $md5 = md5hex $string;
  if (grep { $md5 eq $_ } @uniques) {
    warn "$string is not unique\n";
    # or push() into another list...
  }
  else {
    push @uniques, $md5;
  }
}

# Now @unique contains the list of unique strings
[download]

This should use less memory than what I imagine your solution to be. Note that showing us some of your code can help us give better answers.

Update: Ok, ok. I added the obligatory MD5 :).

Regards.