I found a nice article here.
Basically, you want large sets not in an arrayref, but as a hash, since it's more clear what column holds the data.
e.g.

# tested this on live data, this is twice as fast as fetchrow_hashref,
+ and has the same advantage..
# example taken from the link above.

my @fields = (qw(emp_id first_name monthly_payment)); 
$sth->execute;
my %rec =();

$sth->bind_columns(map {\$rec{$_}} @fields);

print "$rec{emp_id}\t",
      "$rec{first_name}\t",
      "$rec{monthly_payment}\n"
      while $sth->fetchrow_arrayref;
[download]

*Update*
Hmm, Funny indeed, anyway, i still consider this as a valuable way to speed up the DBI. :-)