Article View: comp.os.msdos.misc
Article #4015Re: Detecting Bad Memory Chips of the PC
From: Eric Praetzel
Date: Tue, 27 Jul 1993 14:29
Date: Tue, 27 Jul 1993 14:29
39 lines
3102 bytes
3102 bytes
In article <1993Jul27.132234.8704@rzu-news.unizh.ch> ps@oci.unizh.ch (Sezgey Pachkovsky) writes: >Mike Palmer (mpalmer@encore.com) wrote: >: The power on test does not thouroughly test the memory at all. You need >: to run something like checkit memory test for 24 hours to be reasonably >: certain that memory is OK. >Hope, hope. I had seen machines which will run CheckIt's tests Ok for _weeks_, >but will fail on a simple test of filling each doubleword with it's linear >address and reading it back just once... The problem is what CheckIt tests >memory in 64K-chunks, and if writing to some address will _sometimes_ change Yes you need to know the type of memory chips and their internal design. DRAM chips can fail in wierd and wonderfull ways. Many years ago I wrote a program for fully testing a multi-channel memory buffer and it did the full gambit of tests. Unfortunately rotation tests ... are not much use when you are isolated from the memory by two sets of buffers and some hand-shaking. But then they only wanted to say that we did the full tests, nobody ever said that they worked :-< I had assumed that the power-on tests were some-what thorough. But it looks like they just do a location by location test and a test every 1k if you do the fast pass. Pity. Actually we have some single board computers in one lab that will run every memory test fine. But then for 30 min. or so they will just not run some programs due to memory faults. I almost wish that I had the time to find out what sort of faults the chips were experiencing. Modern chips are getting so large that it is impossible to make them with any sort of yield. Extra memory banks (on the chip) are available and can be switched in if there is a fault with one. This will make thorough testing next to impossible because you no longer have an idea of the structure. Quite a few problems with computers are simply the contacts. Most pins and connectors corrode enough after 6 years to make performance erratic. I have seen new designs do the same thing within a year. Anti-corrosion solutions are becoming more popular to prevent this sort of thing. - Eric
Technical Details
Message-ID:
<CAtuxD.Hpq@watserv2.uwaterloo.ca>Path:
arc1-ca.pugleaf.net!gmd.de!xlink.net!howland.reston.ans.net!math.ohio-state.edu!cs.utexas.edu!utnut!torn!watserv2.uwaterloo.ca!maxwell.uwaterloo.ca!praetzelReferences:
<CAsEyH.A8F@encore.com> <1993Jul27.132234.8704@rzu-news.unizh.ch>Show Full Headers
Complete Article Headers
X-Google-Language: ENGLISH,ASCII-7-bit X-Google-Thread: 10f6aa,4fd176a05271400d X-Google-Attributes: gid10f6aa,public X-Google-Thread: f71eb,4fd176a05271400d X-Google-Attributes: gidf71eb,public X-Google-Thread: f8e37,4fd176a05271400d X-Google-Attributes: gidf8e37,public X-Google-ArrivalTime: 1993-07-27 13:22:03 PST Newsgroups: comp.os.msdos.misc,comp.os.msdos.programmer,biz.comp.hardware Path: arc1-ca.pugleaf.net!gmd.de!xlink.net!howland.reston.ans.net!math.ohio-state.edu!cs.utexas.edu!utnut!torn!watserv2.uwaterloo.ca!maxwell.uwaterloo.ca!praetzel From: praetzel@maxwell.uwaterloo.ca (Eric Praetzel) Subject: Re: Detecting Bad Memory Chips of the PC Message-ID: <CAtuxD.Hpq@watserv2.uwaterloo.ca> Sender: news@watserv2.uwaterloo.ca Organization: University of Waterloo References: <CAsEyH.A8F@encore.com> <1993Jul27.132234.8704@rzu-news.unizh.ch> Date: Tue, 27 Jul 1993 14:29:36 GMT Lines: 39