Subject: Re: true-parity RAM versus psuedo-parity RAM
To: None <port-i386@NetBSD.ORG>
From: Ross Harvey <ross@teraflop.com>
List: port-i386
Date: 05/21/1997 12:52:43
 Alex Barclay adds...
 > Matt Thomas writes:
 >  > At 01:59 PM 5/21/97 -0400, Brian Hechinger wrote:
 >  > >what difference does true-parity RAM versus psuedo-parity RAM make in the
 >  > >PC world?  we had a SUN SPARCstation 2 at work die, and if i act quick i
 >  > >can get 32M of 30-pin 4M true-parity (9 chip) simms, if there is a good
 >  > >reason to do this i'll grab them.
 >  > 
 >  > True parity can do ECC with the 430HX (Triton II) chipset.
 > 
 > Maybe I'm missing something here but surely parity ram will add an
 > extra bit which will catch any single bit error and any odd number of
 > errors but not even numbers of bit errors, and this would truly be
 > parity.
 > 
 > Now if I decide to use more redundant bits say to implement a 4-3
 > hamming code then I can correct any single bit error.
 > 
 > Surely parity is an error detection method, not an error fixing
 > method.

OK, someone has to end this thread! :) "True Parity"==parity. "Psuedo
parity"==logic parity, I suppose. This is a schlock rip-off technique that
substitutes a (cheaper) logic chip for a dram.

Because today's memories are so wide, by the time you get to 64 bits
(simms are installed in pairs these days) you have accumlated 8 parity
bits. These bits, taken as a group, are enough for ECC.  So, using
"parity" simms that have the extra bit-per-byte gives you ECC on good
chipsets that implement it.

Now, Brian was talking about 30-pin old-style SIMMS. I seriously doubt
if any 430HX system board takes these old things. Those simms can't be
used on anything modern. They will work in anything that takes the old
kind, though, because the parity is ignored on boxes that used only
8-bit no-parity SIMMS.
----------------------
Ross Harvey	Avalon Computer Systems, Inc.		  ross@teraflop.com
		Santa Barbara	 		    http://www.teraflop.com