Return to BSD News archive
Path: sserve!manuel.anu.edu.au!munnari.oz.au!uunet!sun-barr!olivea!netsys!news!horizon!sceard!peregrine!ccicpg!mbf!fmt
From: fmt@mbf.UUCP (Mike Trimberger)
Newsgroups: comp.unix.bsd
Subject: Re: Weird Problem (50 MHz, SCSI)
Summary: Beware 50MHz!
Keywords: SCSI, file corruption
Message-ID: <168382@mbf.UUCP>
Date: 16 Oct 92 22:50:04 GMT
References: <1992Oct14.220847.22548@sics.se>
Organization: MAI Systems Corp.
Lines: 36
In article <1992Oct14.220847.22548@sics.se> jonas@sisu.se (Jonas Lagerblad) writes:
>I have had quite a lot of problems with two different 50 Mhz cards, I have
>traced the problem to the cache memory, which I don't think behaves quite
...
>I posted an article to this newsgroup a week ago inquiring if anyone
>besides me have had problems with their 50MHz boards and I have got approx.
>10 answers where most people have had some strange memory problems, most
>of them had a SCSI controller.
...
>My problems have disappeared totally since disabling the cache, but I have
>ordered new faster cache memory, and is eagerly waiting for it to arrive.
We have been evaluating some 50MHz Intel PC systems here and discovered
that the caches don't work. We looked further and discovered that they
were shipped direct from the manufacturer to us with external cache chips
which were too slow to keep up with the 50MH '486. When we went looking
for chips that are fast enough, there were none on the market! Actually
there are a very few but the IC manufacturers are still sampling the new
parts. We were able to get some by having an IC distributer cherry-pick
parts that were labeled slower.
The parts that did NOT work were 12ns rams. We had to use 10ns cache
rams. Only 1 source so far is Toshiba.
>So my advice is to try to disable the cache, and see what happens.
Good advice! If you have problems with a 50MHz '486, disable the
external cache and see if it will work correctly. With the extra
demands that a UNIX system puts on the system, the problems show
up. The reason that our manufacturer never saw a problem, was that
they test the system using MS-DOG and Netware.
--
F. Michael Trimberger MAI Systems Corp.
fmt%mbf.uucp@ics.uci.edu 14192 Franklin Ave
or: {ucivax,uunet!ccicpg,sequent}!mbf!fmt Tustin, Ca 92680
[system administrator in training - sink or swim method] (714) 730-2934