> I'll test that if it will help fix the bug.  But let's exclude hardware
> failure given that I get the same failure with 3 different machines.

There are several types of "hardware failure".  Dead chips (or wires 
or ...) is the one that most people think of when they hear that 
term.  Another one is hardware design bugs.

A third is software errors that show up at the hardware level.  For 
example asking the hardware to do a DMA read from memory that doesn't 
exist. 

Finding the same problem on 3 similar systems doesn't eliminate a 
design bug or a software bug.  (It doesn't really eliminate the dead 
chip problem either, but it does make it pretty unlikely.) 


If other people are running OK with systems similar to yours, the 
obvious place to start looking is the exact hardware configuration.  


Next is the exact software setup.  Did you install 1.4 from scratch?  
(or might there be leftovers from the previous system?)