T20: What Has Failed?

[ArtShapiro]

Time to put on your speculative hats, folks - can you diagnose this one?

My oldest (and first) TP is the T20, which I keep running for no particular reason other than a reluctance to discard it. About a week ago, something happened (I can't remember given all the subsequent grief) that led me to run CHKDSK on it, and there were numerous reports of corruption. I told CHKDSK to fix things, and so many files got diddled that Windows (XP Pro) destabilized.

I decided to just restore a recent image from Windows Home Server, which backs up all of my machines daily.

The resultant restoration wouldn't boot - just a blinking cursor after POST. I've had this happen before and consider it probably to be a Windows Home Server bug - a very infuriating one. When FIXBOOT and FIXMBR from the repair console failed, as usual, to fix the problem, I decided to just reinstall Windows.

A dozen or more attempts to install XP failed - assertions of missing files on the CD, unable to copy various files, etc., etc. I tried three or four different XP CDs and the symptoms didn't change, although the particular files being cited seemed to be inconsistent. I had reformatted the drive several times, quick and full. And the machine seemed to be quite hot, especially around the disk drive.

I enabled the diagnostic POST mode, and discovered that I was only using a quarter gig of memory, not the maxed-out half gig. What a surprise. I swapped the two memory sticks, and was back to a the expected half gig.

I found a CD of the Hitachi Drive Fitness Test. The drive passed the extended check.

I took the CD drive from my T23 and put it into the T20. Tried several more XP installations, with similar poor results.

I took the CD drive and the disk and put 'em both in the T23. I did an XP installation without incident. The machine came up and ran with no difficulties.

I put them back into the T20, and booted. As expected, it blue-screened pretty early due to the different architectures of the two machines.

I then forced a parallel installation of XP on the T20. Each time the installation process complained, I'd push through to retry numerous times, and if need be swap CDs and keep trying. Doing this, all but one DLL eventually got installed.

The resultant parallel installation now runs for a while, but is unstable and ultimately blue-screens.

OK - I think I'm about to run over the T20 with my automobile to put it out of its misery, but I just can't rationalize what component has failed. Disk controller? CD controller? Memory interface? Anyway, speculation is welcome - fire away!

Art

[rkawakami]

I would try three things:

- run a memory diagnostic; memtest86+ from http://www.memtest.org is a good one
- run PC Doctor for DOS on the system; do Systemboard, CPU, Fixed Disk and CD/DVD tests at a minimum (CD .ISO is here: http://www.kawakami-ca.com/ibm_t2x/ibm_t22_pcdiag.iso )
- if those things check out okay, remove the MiniPCI card from the system and try re-installing Windows

[ArtShapiro]

Hmmmm...running Memtest and, while I'm not familiar with its output, it seems to be indicating thousands of errors.

After four minutes, there are over 6000 errors. And the air coming out of the side is quite warm, but that could be due to having the machine physically next to my hand in this brand-new office with a different layout than before.

Have to see if I have any of that old memory sitting around in the junk drawer.

Update: 17000 errors after 10 minutes; everything seems to be in bits 5 & 6 (with bit zero the least significant bit - the manner in which I describe bit locations, which probably isn't the standard PC terminology).

Art

[Neil]

Well, there it is, healthy memory should return 0 errors. Now the only question left is why did your memory fail. Heat? Age? Something else, are all possibilities. I recently had a similar situation with one of my machines, and just tossed the faulty stick and replaced it with another. Now all is well. But, I don't have clue as to what caused it.

[ArtShapiro]

Well, there it is, healthy memory should return 0 errors. Now the only question left is why did your memory fail?

Well, it's getting more interesting. There are two 256meggers in there - ah, the good old days. One is labeled "IBM T20", and MTEST86+ works fine with this guy in either slot.

The other is a Crucial SODIMM, and it fails spectacularly in either slot. All the numbers on it don't seem to "come up" on the Crucial site. But then I looked closer, and it's labeled as a PC-100 DIMM. Their own site says T20 requires PC-133.

Could I have been running the wrong memory all these years (it has a 2002 date code) and had either this DIMM or my T20 aged enough to finally have something go out of spec?

The machine is running obscenely slowly on 256 Megs, but I'm about to reinstall Windows and see if the problems go away. Hopefully I'll find some of the correct memory at home to get this guy back up to a whopping half gig.

Art

[Neil]

The T20 takes PC100 and won't work at all with most PC133 modules. Although Crucial made some low density PC133 modules that do seem to be backward compatible with older machines like the T20. I think the Crucial stick has just failed. Doesn't Crucial have a lifetime warranty?

[ArtShapiro]

The T20 takes PC100 and won't work at all with most PC133 modules. Although Crucial made some low density PC133 modules that do seem to be backward compatible with older machines like the T20. I think the Crucial stick has just failed. Doesn't Crucial have a lifetime warranty?

I was just starting to fill out an RMA form when discovering the assertion (which you believe to be incorrect) that the 2647 T20 takes PC133.

I think, before I go further with the warranty claim, that I'll check my old Dell C610 and see if it uses that same memory. (I think it might have two half-giggers, and the T20 can apparently only take quarter-giggers). I'll confirm that this memory fails in the Dell as well. Then, if they really honor a warranty on something that old, I can feel pretty confident that it's the memory at fault.

Thanks for all the assistance, everyone. I don't think I'll burn the CD Doctor ISO unless something continues to act suspiciously or incorrectly. The repair installation is slowly loading (49%) files as I type.

Art

[Majestic]

Art, after you get it back up and running you might consider monitoring it's operating temperature just to rule out heat as the cause.

Here is a link to a very good utility called Mobile Monitor which will monitor the temperature as well as a few other things....

http://www.softpedia.com/get/System/Sys ... eter.shtml

Given the age of the system, you might take some extra time to freshen up the thermal paste and clean the heat sink of the dust bunnies as a matter of preventive maintenance.

[ArtShapiro]

Update, if anyone cares:

The XP installation went fine - no errors - although slow due to having only a quarter gig of memory. I've applied for an RMA with Crucial. In the meantime, I found a 128meg SODIMM in the junk box, so at least I'm running with 3/8 Gig instead of 1/4 Gig - it helps.

Then I retried the Windows Home Server restoration to that drive. Upon attempting to boot, once again I got the blinking cursor.

I formatted a floppy (arrggh!) as a non-system floppy, and copied BOOT.INI, NTLDR, and NTDETECT.COM to it from another XP box.

Upon booting to that floppy, my XP disk came up fine.

I'm really not sure what this tells me. Clearly the Windows Home Server restoration process has a bug that M$ has never acknowledged. I suspect there's something relatively modest I can do to have the disk bootable without the silly floppy getting involved.