They're similar, but not exactly the same. Putting -w in the command line (or the shebang) will turn on warnings globally, for all the Perl code loaded by that process.

On the other hand, use warnings is a "lexical pragma," and only applies in the block in which it's used. If you put it at the top of a script or module, then it will apply for that entire file (this is called file scope.) Inside a block, it applies only for that block:

#!/usr/bin/perl

print "no warnings out here!" . $undef_var;
{
    use warnings;  # now warnings are turned on
    print "this will warn" . $undef_var;
}

# now warnings are off again.
[download]

(Note: Try running that with and without -w to see the difference.

Typically, it's a good idea to use warnings in your file scope, which allows you to occasionally turn them off in small blocks (using no warnings) if you absolutely must do something that will generate a warning. See perllexwarn for the gory details.

Those are called command-line switches or arguments, and perl's are all documented in perlrun.