The difference is shown by the following example

use strict;
use warnings;
use Data::Dumper;
my @x = (1,2,3);
my $y;

@$y = @x;
$x[1] = 9;
print Dumper $y;

$y = \@x;
$x[1] = 10;
print Dumper $y;
[download]

As you can see, in the second code, the $y is a reference to $x, so they share the same data. In the first case, the $y is a reference to an anonymous array, which has a copy of the data in $x.

Think of the operation @a = @b. This is creating a copy of any array. That is what is happening in the first case (with @$y = @x.

The second case is a $x = \@y, which is assigning a reference to a scalar, so there is no array copying. Only a scalar (the reference to @b) is being copied.