Small nitpicks, making it even easier:

• use the eol attribute to prevent the need for extra print calls
• "<:encoding(utf8)" is much safer than "<:utf8", esp on input
• use autodie to prevent all kinds of error checking (or add auto_diag => 1)

use strict;
use warnings;
use autodie;
use Text::CSV;

my $csv = Text::CSV->new ({ binary => 1 });
my $tsv = Text::CSV->new ({ binary => 1, sep_char => "\t", eol => "\n"
+ });

open my $infh,  "<:encoding(utf8)", "input.csv";
open my $outfh, ">:encoding(utf8)", "output.tsv";

while (my $row = $csv->getline ($infh)) {
    $tsv->print ($outfh, $row);
    }
[download]

otherwise a great reply!