Originally Posted by salasi

I can only offer a slightly handwaving 'explanation'. Presumably one cpu had a microcode bug that wasn't present in the other. Now this is normal and happens all the time, sort of, but usually you either have the bug and the kernel code has to work around it, or you don't and it doesn't. It looks as if your situation in which one cpu has the bug and the other doesn't, doesn't get worked around succesfully. I'm guessing whether the developers thought about it or not, they were unable to do this 'half and half' testing when they wrote the code.

Good that you've found a work-around, though.