I have a GTX 660ti and this config gives me around 216 kh/s
cudaminer --algo=scrypt-jane:UTC -l K7x32 -C 1 -i 0 -m 0 -o stratum+tcp://pickaxe.pool.pm:3306 -u xxx -p xxx
I have a 660 non-Ti and I'm getting 238 khashes for N 9 (MRC/UTC) with -l Y5x32 -H 2 -C 0 -m 1 without lookup gap so you might want to try the Y kernel.