Indeed, it is very inefficient, by comparison to simply doing a linear search. First, you're sorting the list. That's O(N log N) for Perl's default sort type. Next, you're taking a copy of the list (yes, you're copying it) for the purpose of creating the anonymous array. That's about O(N). Next, you're dereferencing the array (a pretty quick action), and then you're popping something off the array (which is also pretty quick, but still is another step).
So what you've got is O( N + N log N ), when you could just have O(N). That's not so good. And as someone else already pointed out, sort @vals does a string sort, not a numeric sort, so 11 will be sorted next to 1 instead of next to 12. If you must use the sort routine, at least change it to doing a numeric sort:
my $max = pop @{[ sort { $a <=> $b } @vals ]};
And here's a linear search for max.
my $max = $vals[0];
$max = ( $_ > $max ) ? $_ : $max foreach @vals;
This version is a little terse, on purpose, to demonstrate that it's possible to do the linear search in just slightly more keystrokes than the sort pop method.
