I wondered if index and substr would be any quicker but, no, it came in a bit behind &_split_LIMIT.

use strict;
use warnings;

use Benchmark qw{ cmpthese };

my $string = q{1:13:1:6:5854:0x00E37F06:0x00D1314C};
my $result;

sub _index {
   my $pos3 = -1;
   $pos3    = index $string, q{:}, $pos3 + 1 for 1 .. 3;
   $result  = substr $string, 0, $pos3;
   }

sub _split {
   $result = join ':', (split /:/, $string)[0 .. 2];
   }

sub _split_LIMIT {
   $result = join ':', (split /:/, $string, 4)[0 .. 2];
   }

sub _regex {
   ($result) = $string =~ m{ \A \d+ : \d+ : \d+ }xmsg;
   }

cmpthese(
   -5, {
   _index       => \&_index,
   _split       => \&_split,
   _split_LIMIT => \&_split_LIMIT,
   _regex       => \&_regex,
   }
   );
[download]

                  Rate       _split       _index _split_LIMIT       _r
+egex
_split        657900/s           --         -26%         -31%         
+-38%
_index        890097/s          35%           --          -7%         
+-17%
_split_LIMIT  952857/s          45%           7%           --         
+-11%
_regex       1066336/s          62%          20%          12%         
+  --
[download]

I hope this is of interest.