Return to BSD News archive
Newsgroups: comp.os.linux.hardware,comp.os.ms-windows.win95.misc,comp.os.os2.setup.misc,comp.unix.bsd.freebsd.misc Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!nntp.coast.net!howland.reston.ans.net!swrinde!gatech!news.mathworks.com!news-res.gsl.net!news.gsl.net!usenet.eel.ufl.edu!arclight.uoregon.edu!netnews.worldnet.att.net!ix.netcom.com!netcom.net.uk!netcom.com!stephenk From: stephenk@netcom.com (Stephen Knilans) Subject: Re: HELP: Can I mix memory speeds Message-ID: <stephenkDuxJLE.EDr@netcom.com> Organization: Netcom Online Communications Services (408-241-9760 login: guest) References: <4s7rae$m3a@symiserver2.symantec.com> <stephenkDutwB2.52D@netcom.com> <31F2C467.167EB0E7@FreeBSD.org> Date: Mon, 22 Jul 1996 05:36:50 GMT Lines: 108 Sender: stephenk@netcom.netcom.com Xref: euryale.cc.adfa.oz.au comp.os.linux.hardware:45163 comp.os.ms-windows.win95.misc:162049 comp.os.os2.setup.misc:17504 comp.unix.bsd.freebsd.misc:24178 In article <31F2C467.167EB0E7@FreeBSD.org> "Jordan K. Hubbard" <jkh@FreeBSD.org> writes: >Stephen Knilans wrote: >> This is a REALLY stupid statement! The chance o a memory error is ALWAYS the >> same, REGARDLESS of the O/S! Also, a crash in UNIX could be so > >I think your own ignorance is showing here, and if nothing else you owe >Joerg an apology for referring to his argument as "stupid" when you, in >fact, are busily making it plain that you haven't got the slightest idea >of what this is all about. First, I am NOT showing ignorance here. Are YOU saying that DOS or VMS or whatever are IMPERVIOUS to memory errors? > >Nobody ever claimed that Parity memory would save the world, simply that >it was a heck of a lot better to KNOW when your memory was going flakey >than not to. With Partity memory, you can at least catch an NMI upon >detection of the error and your UNIX (or any other) OS can then say >"Hey, the state of my hardware isn't looking so hot anymore - I think >it's time to sync the disks and PANIC!!" If I'm administering the >system in question, you can bet that I'll then start swapping memory >modules until those panics go away because the people I build systems >for RELY on them to be stable, and undetected memory errors can only >lead to pain and suffering of the most non-traceable kind. My time is >worth more than that, to say nothing of the time of my customers > >> MANY people have had computers WITHOUT parity or tests, and have had NO >> problems. I had several with BOTH, with NO problems! NOW, though my > >Many people drive without safety belts too and don't necessarily kill >themselves. What does this prove? Do you advocate driving without >belts too? God help us all if you're ever called upon to design an air >traffic control system or a server who's reliability actually counts for >something. > >Needless to say, I can only ask the Impressionable Youth on this >newsgroup to ignore Stephan's advice entirely. While it's certainly >true that your personal box may never suffer from memory problems, >should you ever aspire to creating far more fault-tolerant systems, do >NOT heed his advice! He hasn't a clue and owes us all an apology for so >rudely rebutting the advice of someone who HAS. Parity/ECC is not just >a sales trick, nor is it a panacea. What it is is merely one more >valuable diagnostic aid for those trying to create the most reliable PC >systems it's possible to make with today's technology. >-- >- Jordan Hubbard > President, FreeBSD Project Idid NOT say that parity or ECC, etc... was a "sales trick". Further, I have ADVOCATED ECC(even on THIS group)! There is a BIG difference between parity and ECC. Parity DETECTS errors and aborts. It is *****NOT***** fault tolerant as I generally hear the term. ECC CORRECTS many errors and generally IS fault tolerant. I realize that ECC uses a checksum which is kind of like parity, but it is handled quite differently. I am more careful than you might think. I keep myself grounded when working with MOS based devices, something VERY few people do(though most such components come in bags with BRIGHT stickers that state you should). Also, I have surge protectors on all but my server, and a UPS on my server and needed perhiperals. Need I mention that the chance of a power failure in someway is almost 100% per day(and is probably enough to bring down a computer within 3 years, if only a few minutes due to a power failure). Also, I have many logs, and make sure I have copies of everything. It is funny that you speak of time, new technology, etc.... If you REALLY want a fault tolerant system, get systems with full ECC, hot swap raid (5 or better) drives, full parity(on hardware to enable retrys, remapping) , etc.,UPS, and redundancy. Such a system would be able to stay up through memory problems, CPU, power, disk, etc.... A parity system, by contrast would generally CRASH. I guess, if you intercepted the parity error in the O/S, you COULD block out memory, etc... That, however is bad for several reasons. I HAVE seen people with refresh problems. A parity error in such a case would be just the beginning. BTW Seat belts are nice, but I would REALLY like to have a fast, safe, RELIABLE car with good brakes. Seat belts, if your lucky, only give a better chance of walking away. I'd rather drive away! Steve So, how many people here have parity memory for safetys sake, but plug their systems in the wall, and handle computer boards like a deck of cards? BTW Go ahead Jordan and ignore my advice. The UPS idea ALONE kept the computer at my job running through DOZENS of power failures. HERE, it has kept me up through perhaps a half dozen. As for the handling of cards, I have known people that have had almost 100% failure rate. MINE, is closer to .5%. As for the raid hotswap idea, that would have saved me perhaps 20-30 hours work(I have had a LOT of drives, and they are averaging perhaps 5years TBF). The parity wouldn't have saved me ANY time so far. One friend I had had memory problems, and I quickly found it to be a refresh problem having nothing to do with the memory. BTW, if you get parity errors, or other hardware problems like it, it MIGHT be because you weren't grounded when you worked on the system. I wouldn't be a bit surprised if that caused most of the memory problems.