🚀 go-pugleaf

RetroBBS NetNews Server

Inspired by RockSolid Light RIP Retro Guy

Article View: demon.tech.pc
Article #71415

Re: Memory Parity Errors

#71415
From: John Beardmore
Date: Wed, 29 Nov 1995 00:00
73 lines
3521 bytes
In article <30bba8b2.5624649@news.demon.co.uk> tim-net@nurse.demon.co.uk  writes:

>        Could anyone be so kind as to explain what "Memory Parity Error" means?
>What can be done to avoid them and, by implication, What causes them?

In its simplest form ( which is all most PCs do ), parity gives a check of
memory inegrity by keeping an extra bit per byte of data stored which notes
if an even or odd number of bits were set in the memory write.

If an even number of bits is set when the byte is written, the parity bit
gets set ( or the other way round if it's odd rather than even parity ).

If a byte is read with an even number of bits set and the parity bit is
not set,  PCs that check parity will do an NMI to alert the user and
hang the OS before much damage can be done to data. This sounds really
unfriendly I know, but if you've ever spent 3 weeks with a PC that has an
obscure memory problem and doesn't report it, you'll realise just how
excellent parity checking is.

Parity checking is done in hardware as the process must take only a few
nanoseconds.

One thing to watch for. Some simms only fake parity. They evaluate the
value of the parity bit as data is read and thus don't detect errors.

Of course, parity checking has one problem. It detects single bit errors
but not errors where an even number of bits get corrupted. As single bit
errors are far the most common, this is not much of a problem.


Generally RAM errors fall into two categories:

  Soft errors due to things like vibration or the radioactive decay of
  contaminents in the slicon used to make the RAM.

  Hard errors that always occur when a cell in the DRAM matrix goes poot.

Things like errors created by RAM that is being used on the edge of its
thermal or spede envelope kind of fal between the two stools; they are
reproducible but they do not indicate disfunction so much as impropper
use !


Avoiding parity errors:

  Don't just turn of parity checking !!  Find out what is wrong or you
  will suffer horrible problems in the long run.

  First check that the simms are seated correctly; if they are not you
  will see errors. Incorrect mechanical contact is probably the biggest
  problem.

  Make certain that your simms are metalurgically compatible with the
  sockets. Gold plated simm contacts are not said to get on well with tin
  plated sockets as far as I recall.

  Make certain the memory is fast enough for the machine and that the chip
  set is inserting enough wait states in RAM access cycles.

  Make certain that the machine runs cool. Fans must work, ventilation
  slots must not be blocked, watch for the holes in the PSU case getting
  fluffed up !


If you still have problems, you can try using less memory than normal,
then swapping simms from bank to bank to see which is at fault. If that
fails to nail down the problem then getting somebody with a simm tester
ought to be a better test.


Let us know how you get on,  J/.
-- 
John Beardmore
Technical Details
Message-ID: <817672737snz@wookie.demon.co.uk>
Path: arc1-ca.pugleaf.net!unknown!not-for-mail
References: <30bba8b2.5624649@news.demon.co.uk>

Show Full Headers
Complete Article Headers
X-Google-Language: ENGLISH,ASCII-7-bit
X-Google-Thread: 1036bc,5d944687e07667e8
X-Google-Attributes: gid1036bc,public
From: John Beardmore <wookie@wookie.demon.co.uk>
Subject: Re: Memory Parity Errors
Date: 1995/11/29
Message-ID: <817672737snz@wookie.demon.co.uk>
X-Deja-AN: 120411547
x-nntp-posting-host: wookie.demon.co.uk
references: <30bba8b2.5624649@news.demon.co.uk>
organization: Myorganisation
reply-to: wookie@wookie.demon.co.uk
newsgroups: demon.tech.pc