Date: Mon, 6 Nov 1995 10:54:54 -0800 (PST) From: Rob Mallory <rmallory@wiley.csusb.edu> To: current@freebsd.org Cc: smace@metal.ops.neosoft.com Subject: Re: vnode_pager_outpages error Message-ID: <199511061854.KAA07125@wiley.csusb.edu> In-Reply-To: <199511051946.LAA12769@GndRsh.aac.dev.com> from "Rodney W. Grimes" at Nov 5, 95 11:46:02 am
next in thread | previous in thread | raw e-mail | index | archive | help
Here are a few notes from my last two months tracking down my problems on a 4xe/pb cache/100Mhz 32MB in 2 16MB 60ns simms: [...] > simply clock your 100Mhz chip down at 90 (change external clock from 66 to > 60Mhz). Didn't help. 'just had to wait longer for a panic. > > I've tried it with -current and -stable and still get signal 11's but less > > frequently with -stable. same here...I went a step further and dropped the PHK-Malloc'd libc into stable and found this hi-po malloc to aggravate the problem...[compliment to Pohl on his tight and efficient code;)] > The #1 cause I have found of sig-10 sig-11 that are _hardware_ related have > to do with memory system failures, quite often a bad cache chip or bad cache > module and/or bad main memory. You _are_ using 60nS main memory right??? > You do have 24 or less chips on each simm, right?? EIther of those violate > the stated operating parameters of the PCI/I-P55TP4XE board. One thing I just found about, and have yet to see in print, is that there are different grades of chips with faster and slower 'refresh timings'. More specificaly, there are "universal" (2k refresh) and "mac/triton" (4k refresh) simms which are not interchangeable. (2k will work anywhere but not <-->) 4k simms (designated by xxx164xxx on the chips themselves) 'should' work in triton chipsets, but from what I have seen (siemmens and a mix of samsung and ti-japan) ..they do not. supposedly, the samsung's are making a comeback, and are the only ones which do work, (at 4k refresh) but i only had one. The "2k refresh" simms are designated by xxx174xx. I will hopefully be getting the "good ones" this week for my 2-month-old MB. It can be hard to put the blame on RAM with all the strange errors that bad, non-parity ram is capable of producing. -you keep looking elsewhere because of a mental-block about not seeing "parity error on xxx" Take this with a grain of salt, (it involves win95 and corel-draw) but what made me "beleive" (after swapping cpu,cache,ncr for known good ones) was after playing with CorelDepth-6 rendering a 16MB 3D logo, was being able to "see" the faulty lines in memory. patterns (usualy diag rows of lines) would pop up in the rendering as pixels changed from some color to white and the upon re-rendering, or transforming it, they would move. it never really crashed, at all. ...just think if microsloth used the PHK-malloc! people would run into the streets screaming that win95 is full of bugs, while its really all the low-grade simms imported from shoe-factories in Tiwan. > Phk just flashed his machine up to that revision and is now seeing soft > (C-A-D) reboot problems, any such things going on with yours? I checked > here and can not duplicate it here. Make sure you use option#3 and update both halves of the rom, and also 'load bios defaults' after flashing; let it boot until you see 'updated ECSD successfully' or whatever, then reboot and play with bios values. ..this was a big one on the -asus lists for a while. Rob Mallory [mallorrp@sce.com][rmallory@csusb.edu]
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199511061854.KAA07125>