Date: Mon, 23 Nov 1998 20:52:36 -0500 (EST) From: Chuck Robey <chuckr@mat.net> To: John Polstra <jdp@polstra.com> Cc: "Jordan K. Hubbard" <jkh@zippy.cdrom.com>, current@FreeBSD.ORG, rom_glsa@ein-hashofet.co.il, Poul-Henning Kamp <phk@critter.freebsd.dk> Subject: Re: Random craches under heavy(?) disk activity Message-ID: <Pine.BSF.4.05.9811232049270.302-100000@picnic.mat.net> In-Reply-To: <XFMail.981123174143.jdp@polstra.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, 23 Nov 1998, John Polstra wrote: > On 24-Nov-98 Jordan K. Hubbard wrote: > >> I would actually argue that the longer the box has been overclocked > >> the warmer the cpu runs. > > > > I've also had OC'd boxes start failing. A trusty PPro 233 of mine ran > > OC'd for almost 2 years before finally starting to exhibit occasional > > flakiness. I now run it at 200 and it's happy again. > > I wouldn't trust it even back at 200. Overclocking causes > overheating, and it's not uniform overheating. Every chip has hot > spots which get much hotter than others. Research has clearly shown > that such localized overheating damages the chips, makes them less > reliable, and shortens their lifetimes. > > I used to do a lot of consulting work for a company that built > equipment for diagnosing problems in digital circuit boards. The > goal was to find the bad chip on the board, and the essence of the > process was to somehow inject test patterns that would stimulate the > inputs of individual chips in controlled ways, and then see whether > their outputs matched expectations. We investigated a technique > called overdriving. The idea was that you used high-current drivers > to just blast the test patterns into the chips' inputs, very briefly > overwhelming the outputs of the chips that were driving those inputs. > The hope was that we could keep the duration of the overdriving brief > enough so that the other chips wouldn't be damaged. We abandoned > it eventually. Even the shortest useful periods of overdriving were > enough to reduce the lifetimes of the chips being overdriven. > > I believe that similar damage most likely occurs when you overclock a > CPU. It doesn't mean a thing that the package feels cool. There can > still be tiny spots on the chip that are badly overheated. An instructor of mine tells me that you can actually use a process that's kinda like silicon sandblasting, to leach away a small part of a chip, and literally look at failure modes. Something like very high energy sputtering ... Intel has used it to uncover bugs that would have taken months to discover using other analytical techniques. I'll have to ask him now about the failure scenario you're painting. > > John > --- > John Polstra jdp@polstra.com > John D. Polstra & Co., Inc. Seattle, Washington USA > "Nobody ever went broke underestimating the taste of the American public." > -- H. L. Mencken > > To Unsubscribe: send mail to majordomo@FreeBSD.org > with "unsubscribe freebsd-current" in the body of the message > > ----------------------------+----------------------------------------------- Chuck Robey | Interests include any kind of voice or data chuckr@glue.umd.edu | communications topic, C programming, and Unix. 213 Lakeside Drive Apt T-1 | Greenbelt, MD 20770 | I run Journey2 and picnic (FreeBSD-current) (301) 220-2114 | and jaunt (NetBSD). ----------------------------+----------------------------------------------- To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.05.9811232049270.302-100000>