mike smith writes:
> Well, some very interesting results were obtained; sorry about
> taking so long to get back, but I've been through the build
> cycle about four times now in order to try to confirm my 
> suspicions.
> 
> A recap : my overclocked 425t was failing mysteriously during 
> 'make build' with spurious but repeatable errors (compiler 
> signals, garbage in intermediate files, etc.)
> 
> Building at 25MHz worked fine, confirming Jason's suggestion
> that the overclocking was at fault.
> 
> Building at 33MHz works fine as long as I remove the 80ns
> RAM cards, leaving only the 70ns parts.

This makes sense (for chips like 680x0 and i486)

[anecdote for i386 users]

For Pentium class (and higher) machines, any CPU that runs at a
multiple of 30MHz uses 70ns RAM; to overclock to a 33MHz multiple
you must use 60ns or faster. Once I had upgraded my system to 60ns
RAM I then overclocked my P5-90 to P5-100 without problem.