A Magical mystery tour – John Saxon

I have finally recovered sufficiently to be able to talk reasonably coherently about a recent "Weekus Horibilus". Actually I was without my system for almost 10 days – talk about withdrawal symptoms! I was very lucky to have friends who rallied round, lent me lap-tops and various components in attempts to fault find by substitution, but (slap my wrist!), I was actually close to seeking professional help. Psychiatric and technical!

As usual with me, there was an element of self infliction, and as this fault ended as being rather obscure, I thought I would set the hardware/software pundits a test. Can you solve the mystery before the final disclosure? Do not skip ahead!

A gentle start:

It all began with a trip to the markets where a new hard drive (6.4gb Quantum fireball) was purchased, plus a cheapie modem (Xlink 56Kb K6Flex – with floppy to flash to V90). My system (running W98) is a 200mhz AMD K6 MMX on a Tyan Tomcat HX MB, 48Mb RAM with an 850Mb Connor C: drive, and a WD 1.6Gb D: drive + most of the usual other peripherals. The motherboard is rather large with 8 SIMM slots, 4 of which are almost out of sight (and reach) underneath the hard drive bay. Some time ago, to provide better access, I drilled out the drive bay rivets and replaced them with screws and bolts. As I was also fitting a removable drive bay for backup (to use the replaced Connor 850mb drive), this seemed like a good time to do some mechanical and cable rearrangement. So out came the drive bay and the cables, and as I could now get at the motherboard, the SIMMs, various ICs, and jumpers were given the obligatory re-seat pushes. My experience is that at least 75% of hardware faults are contact related. Everything was then re-assembled and re-connected. The new 6.4Gb drive was installed as the master drive on the primary IDE channel, powered up and (using the W98 emergency disk ), partitioned 4 ways (3 X 2Gb and the rest - all FAT32). Then it was formatted with system files on the first active partition. But…. NO BOOT from the new C: drive. BIOS recognised all the drives OK, POST checks were good and DOS diagnostics ran well, but the boot sequence hung at "Verifying DMI pool data". The thought crossed my mind that perhaps W98 would only boot to DOS via the F8 trick, or perhaps DOS did not like FAT32? So I pressed on with cloning W98 system from the old to the new C: drive.

The downward slope:

A slight rearrangement of drives took place – the WD 1.6Gb was disconnected and the setup changed to the Connor 850Mb as prime and the new Quantum as slave. It should be mentioned at this point that the Connor has given me trouble before. It’s master /slave jumper labelling is horribly ambiguous. So it was no real surprise when the system would not boot – in fact it would not even complete the bios checks or allow entry to the CMOS setup – POST tests and video were working but almost nothing else. The only solution (at the time) seemed to be to return the CMOS to default parameters – but the CMOS could not be accessed as the BIOS was not completing. So with great trepidation and gritted teeth, the motherboard jumper was shorted for more than 10 secs (per MB documentation) and the BIOS then ran and allowed the drive jumpers to be sorted out.

Now a new piece of software was tried – DriveCopy by Powerquest – the people who wrote Partition Magic. This is advertised as "the fastest way to clone your system to a new hard drive". It turned out to be really fast – more than 40mb/minute, but a little too automated for my taste. It wanted to rearrange all my carefully installed partitions, so I decided not to use it. It should work well with a brand new unformatted drive however.

So back to good old reliable Ghost clone software (described in a previous 16Bits article). Cloned the old C: drive to new E: partition (for safety) then tried "Old C: to New C:" but ghost aborted with "bad clusters on the source drive".

At this point the new drive was tested on a friend’s machine (many thanks Merv) and it was confirmed to be faulty – it should boot without W98. So I decided to return to the old C: drive until the Quantum drive could be replaced (had to wait till the next markets in 2 weeks). As Ghost (and Scandisk) said that the old C: drive had bad clusters it was decided to re-format because a system backup was available on the 6.4Gb E: partition. WRONG! Cloned E: to C: and disaster! Corrupted registrys, cross linked files, bad FAT tables, etc. Nothing left but to re-format and install a "clean" W98 system, and look forward to weeks of re-tuning. So re-formatted C: /S and system booted to DOS O.K.

From Bad to Worse:

Then tried to do the easy part – a clean W98 install. But the process halted (no error message) during "preparing setup files" During the next attempt, Scandisk reported many file and FAT table errors – corrected these but the install halted again at a different point (again no error messages). Again Scandisk reported numerous corrupted files.

I am still trying to forget the next few days – sheer misery! I had a great DOS system – could copy software between drives and run it quite O.K. But after numerous re-formatting and install attempts, nothing I could think of would persuade it to install W98! How are the hardware pundits doing?

Again with the help of friends, I tried a replacement CDROM drive, replaced cables, replaced disk drives, and tried other W98 install disks. Even went out and mowed the weeds – anything to avoid the inevitable conclusion that there was an obscure CMOS or serious Motherboard problem. I did not wish to know that!

The road to enlightenment:

Nearly 10 days without my system! Luckily I was able to borrow a laptop and modem – "users helping users" again (Many thanks Gloria), so I was not totally cut off from the virtual world.

By this point I was seriously considering paying someone to fix it – what a come down. But I needed someone familiar with AMD K6s as I still had a sneaking suspicion that there was some obscure fault relating to the CMOS "defaulting". I thought I remembered that the K6 needed some "special" CMOS setting – but I had tried every CMOS parameter combination that seemed even half way reasonable.

Then I ran into Darrell Burkey – an ex-editor of this august journal and a very knowledgable type. I poured out my tale of woe, and after discussing the possibility of re-flashing the CMOS, he thought for about 30 secs and casually said "why don’t you check out upper memory"?

This was a real "forehead slapper" for me. OF COURSE! Without High Memory drivers etc., DOS only uses the bottom 640Kb of memory – probably the first time the rest of memory is used is by Ghost and other clone programs, and specially by the W98 install software! Now if there was an intermittent fault in upper memory…..

And "the rest is history" as they say. I had run simple DOS diagnostics (including memory diagnostics) which had passed O.K. But I needed something (up to date) to really wring out upper memory. Luckily I also had previously downloaded BCM Diagnostics which are a very nice set of diagnostics from http://www.bcmcom.com. They include some DOS diagnostics and sure enough they indicated a fairly intermittent upper memory problem. I must have caused this when I gave everything the magic re-seating pushes. So out came the SIMMs, a quick edge connector clean with alcohol, popped them back into different (more accessible) slots, and the diagnostics were perfect over a considerable number of runs.

After this W98 installed perfectly and so did all the applications, etc.

Lessons learned and postscripts:

    1. After going around in circles for a while – give it a rest and ask other friends. One can get much too close to some problems. A fresh mind can often help.
    2. When buying from the markets, bear in mind that although you may save quite a lot of money, you may have to wait quite a while for a replacement. So "mission critical" components need extra consideration.

"And what about the new Modem?" I hear you ask. You did, didn’t you? Well there was a minor tragedy there as well.

It ran OK in K56 Flex mode, but the "flash to V90" floppy disk was nagging me – another challenge! The flash program first identifies the modem and port (I think it uses the W98 parameters), then after the user give the go-ahead, it downloads a loader program into the modems flash memory, then that program loads the new modem software. There is a point during this process where the loader has been installed but no operating software. At that point you do not have a modem – just a bunch of components that do not know they are supposed to Modulate and Demodulate.

You have probably guessed what happened to me . Yup – the flash software stopped after identifying the modem and downloading the loader. At that point the modem cannot be recognised by W98 or the flash software, and so no more flash attempts can be made. There is no "force a flash" mode. What a crazy system – no way to back out! So the device has to be sent back to the manufacturer. I am happy to say that the dealer agreed to exchange the Modem as well as the Quantum drive. So now I have a replacement K56 Flex modem and am in no great hurry to flash it to V90.

P.S. The replacement drive is working well – but again the old Connor drive let me down. Ghost reported that my laboriously rebuilt system on the C: drive had bad clusters when trying to clone to the new drive. So another two weeks of system and application installing and tuning after another "clean" W98 installation.

Are these computers working for us? Or are we working for THEM?