Just want to let you know the results of my optimization:

4x XFX 5870 1GB
LinuxCoin
Clocks: 900 / 300, standard voltage (AMDOverdriveCtrl)

phoenix.py run from command-line (not start_mining script)
kernel: phatk (15-20 MHash/sec higher than withpoclbm)

flags:
VECTORS
AGGRESSION=9 (Increasing AGGRESSION makes system slower and does not further increase hashrate)
BFI_INT
WORKSIZE=256 (10MHash/sec higher than with 128)
DEVICE:  (0,1,2 or 3)

hashrate: 395 MHash/sec
temps: 63-78 ^oC (lowest to highest mounted card)

Cards are on a 890FXA-GD70 with a 1200W CoolerMaster PSU in a closed HAF X at 25 ^oC ambient temp.