LM3886 gain clone vs STK4048XI for 4 ohm load

[filgor]

I am questioning wheather I should use a single STK4048 or three LM3886's in gain clone parallel (each seeing 12ohms) to drive a 4 ohm (2.83+2.83j semi-reactive) load.
I will be Using a heatsink rated at 0.74degC/W

Has any one done a real analysis on the STK4048XI using regulated power supplies and or 4ohm loads?

The datasheets available from Sanyo for the STK range are somewhat opaque compared to the National semiconductor datasheets for the LM3886.

What I am resonably sure of from the available data is that the three LM3886's in parallel would provide lower distortion but would have a higher thermal resistance as follows.

Distortion:
0.03% at 0.1 watt at 20KHz 8ohm for STK4048XI from data sheet'
0.02% at 0.1 watt at 20KHz 4ohm for 2 x LM3886 in parallel (each seeing 8ohm).
one could only assume the STK4048 is worse at 4 ohm and that 3 LM 3886's in parallel provides lower distortion than 2.

Thermal resistance:
1.4/3=0.47deg.c/W for 3 x LM3886 as apposed to 1.2/4=0.3deg.c/W degrees c/W for the STK4048... MUCH better!

The other factors I am aware of are as follows:

1. The LM3886 has built in thermal shutdown

2. Maximum supply voltage. The LM 3886's will only tolerate +/- 42 volts while the STK will take a whopping +/-87 volts... BIG difference!

3. The convinience of one STK4048 IC as appossed to three LM3886's with 3 sets of matched feedback networks and output resistors!

4. In Australia one STK4048XI is much cheaper than 3 LM3886's

Has any one got any information that is contrary to any of the above or that would be useful in my decision?