From owner-freebsd-hackers Sat Oct 25 15:57:21 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id PAA28601 for hackers-outgoing; Sat, 25 Oct 1997 15:57:21 -0700 (PDT) (envelope-from owner-freebsd-hackers) Received: from out2.ibm.net (out2.ibm.net [165.87.194.229]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id PAA28594 for ; Sat, 25 Oct 1997 15:57:18 -0700 (PDT) (envelope-from mouth@ibm.net) Received: from slip129-37-53-68.ca.us.ibm.net (slip129-37-53-68.ca.us.ibm.net [129.37.53.68]) by out2.ibm.net (8.8.5/8.6.9) with SMTP id WAA76996 for ; Sat, 25 Oct 1997 22:57:10 GMT From: mouth@ibm.net (John Kelly) To: freebsd-hackers@FreeBSD.ORG Subject: Re: Parity Ram Date: Sat, 25 Oct 1997 22:58:33 GMT Message-ID: <34536bc3.4216043@smtp-gw01.ny.us.ibm.net> References: <34524948.41C67EA6@est.is> <34525F3B.1137B612@ix.netcom.com> In-Reply-To: <34525F3B.1137B612@ix.netcom.com> X-Mailer: Forte Agent 1.01/16.397 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 8bit X-MIME-Autoconverted: from quoted-printable to 8bit by hub.freebsd.org id PAA28595 Sender: owner-freebsd-hackers@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk On Sat, 25 Oct 1997 17:06:03 -0400, Jerry Hicks wrote: >Didn't Richard Hamming show these to -cause- more problems than they >solve? It seems I recall a number like 256K (bits/bytes/words?) as being >the threshold in a proof he presented. Even if he's correct that additional bits needed for parity increase the overall frequency of bit errors, when a one-bit error does occur on a parity SIMM, at least you are notified of that fact. On the other hand, any bit errors occurring on a non-parity SIMM will be unreported. You will have corrupted memory, which could be as trivial as a reversed bit in a graphic image, or a serious data error. The fact that you can change one bit in a graphic image without dire consequences is why printers don't need parity memory. But on the other hand, most banks use mainframes with ECC memory. Would you be a satisfied customer if they kept your account balance on a PC with non-parity memory and every once in a while subtracted $16,384 from your account because they had a bit error in memory which went undetected? So it depends on the application. Non-parity memory has its place, but not on any computer where data integrity is important. Intel published a white paper which claimed that with modern memory manufacturing techniques, bit errors due to gamma radiation and other disturbances are no more likely than once every 20 years or so for a DRAM chip. But that's just for one chip --- multiply the number of SIMMs and individual chips found in a machine and the likelihood increases. A great shame upon the computer industry is that chip makers like Intel have foisted non-parity chipsets like the 430FX upon an unsuspecting and uninformed public who imagine their PCs operate reliably as any other appliance. With so many non-parity chipsets and memory in use, running bug-ridden Microsoft products, I'm amazed the American economy hasn't collapsed entirely. John