Re: time_t size: The year 2038 bug?

[Ludovic Aubry]

"Jakma, Paul" said :
> >       Not so. The 64-bit chip can run with a lower clock rate
> > and still get the
> > same amount of work done, reducing EMI. Why will it need more or more
> > exensive ancillary compents? No law says the external bus
> > implementation has
> > to be 64-bits.
> >
> 
> 64 bit chip needs significantly more gates than a 32bit one. and hence will
> draw exponentially more power, unless it's made with a significantly higher
> tech process -> which means cost. Bus interface is another factor.
> 
> old chips have a minimal (even free) design cost. Fab is cheap cause you can
> use an out-of-date process. Yield is very high because the process is well
> tested and understood.
> 
out-of-date process may not be cheaper, because you may well not find a
factory using it, on the other hand,
with a given process, whatever the chip, the waffer cost is constant,
the difference is whether you put 1000 4-bits chips on a 100$ waffer or
10 64-bits chip.
moreover if you get, say 10 errors (because of dusts, or whatever) on
your waffer, you lose 1% of your production in the first case, and 100%
in the second case (if you're unlucky enough so that you have one error
on each die).
Basically that's why bigger die size costs more than smaller ones.

(PS: I'm not sure of the spelling (maybe wafer) anyway I mean the big
round plate of silicium where you engrave your chip.)
regards,
Ludovic Aubry