From owner-freebsd-questions  Mon Jul 14 05:47:20 1997
Return-Path: <owner-questions>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.5/8.8.5) id FAA19074
          for questions-outgoing; Mon, 14 Jul 1997 05:47:20 -0700 (PDT)
Received: from implode.root.com (implode.root.com [198.145.90.17])
          by hub.freebsd.org (8.8.5/8.8.5) with ESMTP id FAA19062
          for <questions@FreeBSD.ORG>; Mon, 14 Jul 1997 05:47:07 -0700 (PDT)
Received: from implode.root.com (localhost [127.0.0.1])
	by implode.root.com (8.8.5/8.8.5) with ESMTP id FAA06163;
	Mon, 14 Jul 1997 05:47:10 -0700 (PDT)
Message-Id: <199707141247.FAA06163@implode.root.com>
To: Nadav Eiron <nadav@barcode.co.il>
cc: dmaddox@scsn.net, Wes Peters - Softweyr LLC <softweyr@xmission.com>,
        Nick Johnson <spatula@gulf.net>, questions@FreeBSD.ORG
Subject: Re: A few solutions 
In-reply-to: Your message of "Mon, 14 Jul 1997 15:02:15 +0300."
             <33CA1547.AD3@barcode.co.il> 
From: David Greenman <dg@root.com>
Reply-To: dg@root.com
Date: Mon, 14 Jul 1997 05:47:10 -0700
Sender: owner-questions@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

>No. What you need is ECC motherboard and ECC RAM. To be able to correct
>memeory errors you need more bits than what's available on a parity
>SIMM. Some ECC implementation (the one I have in mind is the AlphaServer
>1000, don't know if Pentium MBs have this too) use standard RAM, but in
>greater quantity. The AlphaServer 1000 has banks of 5 standard SIMMs,
>instead of the 4 that would otherwise be required for its 128 bit memory
>bus. It uses the extra memory to implement ECC. Later models used 4 ECC
>SIMMs instead.

   ECC for 64bit words requires 8 syndrome bits. Coincidently, that just
happens to be the number of parity bits you'd have if you had byte parity
SIMMs, so modern PC motherboards use the 8 parity bits to implement ECC
and no special ECC SIMMs are required.

-DG

David Greenman
Core-team/Principal Architect, The FreeBSD Project