MIKE ROSS wrote in a message to All:

 MR> Hi,

 MR> Could someone tell me why parity checking has been dropped from 
 MR> bios?

Because it's cheaper.  I'm not real sure of the timing of this stuff,  but
I suspect that it probably happened when ram prices started getting way up
there,  for what was considered "required" ram in a system
whenever that was. THere was a time or two...

 MR> I still have an award 486 bios which has parity on the dram. Isn't 
 MR> parity checking still something that's needed?

I would think so,  probably more so now than it was back when.  Memory is
much denser,  and the cells are much smaller,  and that would seem to make
it more likely to have intermittent parity errors because of bits read
wrong.

 MR> If not then why did it seem so important for so many years. I 
 MR> can't count the times a parity error would stop the system dead 
 MR> only to find a lifted dram or so. 

Lifted?  As in out of a socket?

Frankly,  given the choice,  I think I'd prefer ECC memory,  which uses
some more bits to _fix_ the problems.  I read somewhere that memory gets
about one "soft" error a month,  which I suspect most software
(including operating systems!) ignores,  sometimes with little effect, 
sometimes crashing spectacularly.  A hard error *should* stop the system, 
but soft errors should be corrected and the system allowed to continue
running,  as stopping a multi-user multitasking OS in its tracks is not
generally a good thing.

But then,  my ideas of what's good hardware design and that of industry
don't seem to be in agreement,  and haven't been for some time now. 


---