Re: Ram lost.

Chad Page (
Mon, 3 Mar 1997 16:37:40 -0800 (PST)

On Mon, 3 Mar 1997, Egor Egorov wrote:

> Yet another trouble - I have 32mb memory, but 2.0.29 detects only 15. If I
> say mem=32M to lilo, gcc does sig11 everytime I run it. Well, seems to me,
> there is a hole in the memory at 15mb. What to do ?
(A minor possiblity is that the kernel is not configured for
systems > 16MB... the more likely one is what I mention next...)

Sounds like bad memory - sometimes when memory really goes bad
(e.g. a lot of Fry's Electronics so-called 'premium'* simms for those who
live near San Jose) the BIOS won't even bother to recognize it. (It's
better than the alternative, which is a bunch of sig11's...) I know
someone who tried to use Fry's memory on a Micron with a Micronics
HX-based board and the BIOS didn't detect it... (And those were 'new'
simms although Fry's has the habit of putting returns up for sale
without testing them first...)

(* - The Toshiba SIMMS BTW are generally much better, and have
lifetime warranties. The flip side is that other stores often have
quality memory for a much better price... in general ya' get what you
pay for :)

The real problem is when the memory gets detected but has
problems on a motherboard like the old Triton FX boards which can't use
parity, or systems with EDO memory (thanks to the FX boards which made
EDO popular, there are almost no parity EDO simms). At least with parity
or ECC a less experienced user will be able to know for sure if it's the
memory or not and can adjust memory timings and/or replace simms
accordingly depending on the problem. I've gotten a 'bug' report on the
ramdisk driver caused by someone with ~224MB of non-parity memory - and
one of the simms became flaky and corrupted kernel buffer-cache tables. Ouch.

Sorry 'bout the Fry's rant earlier, but I suspect there's a store
like it near most of us on the list. Unfortunatly.

- Chad,
(parity/ECC memory advocate :)