Robert Bruce Thompson
|It now appears that Robert Bruce Thompson and I
will be doing a book on "good enough hardware." Stay tuned. Thompson is the
author of several O'Reilly books. I have never seen a bad O'Reilly book, which is
Other reports by Thompson:
An Upgrade for Old Kerby
by Robert Bruce Thompson
As regular readers of my Day Notes journal will recall, I've been intending to upgrade old kerby (a Gateway 2000 486/66 tower system) for a month or more, and been serious about it for the last couple weeks. Finding the time has been the problem, but last night (Sunday, September 20, 1998), I finally got around to doing it.
Originally, I'd planned to do a fairly minor upgradebuy a Pentium/200 system board or something similar, swap in the 96 MB of 72-pin memory from old kerby, and have done with it. The first problem was that I was mistaken in thinking that kerby had an ATX system board and power supply. It didn't. Although the keyboard and mouse connectors are PS/2 style (which often indicates an ATX), kerby in fact had a baby AT system board and power supply.
I wanted to use an Intel system board, if possible, but Intel really just makes system boards in order to sell processors, so about all you can get is whatever holds the current processor Intel wants to sellthe Pentium II in this case. I had a Pentium II sitting on the shelf from my aborted attempt to build a dual-processor box on a Micronics Helios dual-processor system board. I ended up sending back the Micronics, swearing never to purchase another of their products, and installing an Intel Seattle single-processor board instead. That left me with the PII sitting on the shelf.
When I got to Computer &; Software Outlet, I found that Intel wasn't a possibility unless I wanted to buy a new case and power supply. I ended up buying an EPoX EP-68LXR system board, two 64 MB DIMMs (can't reuse the 72-pin SIMMs on a PII system), and a bunch of other stuff. Kerby's "minor" upgrade turned out to be one of those "jack up the radiator cap and roll a whole new car underneath it" deals. No big deal. Barbara's still using an old Gateway Pentium/133 tower system, so I figured I'd build a new PII system for her. I ended up spending about $650 on parts (not counting the processor or video card, which I already had on the shelf). That's not bad for a PII system with 128 MB RAM, even counting the parts I planned to recycle.
At any rate, I really need a system running Win98 for the book I'm writing currently, so I decided to take the time to get the upgrade done last night. I spent half an hour before dinner tearing old kerby down, removing the old system board, and blowing out the dust bunnies. All the front panel jumpers were labeled on the system board end, so that's one step I didn't have to worry about. After dinner (a can of chicken noodle soupI really wish my wife would get home from her tour of Canada) I got down to business.
Configuring and Installing the System Board
I installed the PII processor and one (of two) 64MB DIMMs on EPoX EP-68LXR system board. If you're going to fail the smoke test, there's no point in burning up more than you have to. The EPoX system board has a jumper that allows you to set bus speed for 66, 68, or 75 MHz. It's jumpered by default for 68 MHz, which sounded a little hokey to me, so I removed the jumper to set it back to 66 MHz. I've never been a believer in overclocking CPUs and other components. The CPU speed is set by DIP switch, which I set for the 300 MHz Pentium II.
After doing a quick visual examination to make sure the mounting holes lined up, I started to actually install the system board. The first problem was the I/O template on the back of the case, which had cutouts for PS/2 mouse and keyboard connectors. The EPoX has a PS/2 mouse connector on system board, but has a standard AT keyboard socket poking out the back in the traditional position.
The sheet metal on the Gateway case itself has a cutout that will fit the AT keyboard connector, but the thin metal template wouldn't allow the AT connector to clear. I tried to remove the template neatly, but the only way to get it out was to use brute force. I used an old screwdriver to bend and twist the template until it popped out. That left me with a hole about 1/2" by 2", from which the AT keyboard connector on the EPoX protrudes nicely.
I suppose that in theory I'm now in violation of some FCC RF emissions regulation, but I'm not going to worry about it. I'm more concerned about air flow. The PII runs pretty hot, and although it has a cooling fan mounted directly on it, I don't want to do anything that damages the air flow around it.
Connecting the Cables
Well, theres a 5-pin header for the PS/2 mouse connector on the system board. Pin 1 is labeled, but nowhere in the docs does it say how to orient the matching connector on the motherboard jack end of the wire that has the PS/2 connector for the back panel. That connector is a 5-pin female jack, with wires colored in order, yellow, <nothing>, green, blue, red. If I had to guess, Id put yellow at pin 1, but fortunately I hit the EPoX web site and they had just what I needed. Pin 1 is red, pin 4 n/c, and pin 5 yellow. Exactly the opposite of what Id have guessed.
The next problem is the connectors for the serial and parallel ports. Old kerby already had three connectors on the back panela DB25F for parallel, a DB25M for Serial A, and a DB9M for Serial Bthat connected to header pins on the old system board. The new system board uses exactly the same arrangement, and comes with its own set of the same three connectors. The trouble is, theyre on two expansion card slot covers, and I cant afford to waste those slots, particularly since the PS/2 mouse port is already taking one expansion slot cover.
No problem, I figured. Ill just use the existing back panel connectors already on old kerby. The trouble is, the header pin connector jacks on those adapters are keyedthey have one hole filled to make sure you cant insert the jack backwards. The male header pin connectors on the old system board have the proper pins missing. The ones on the new system board are fully pinned, so the existing connectors wont fit. Okay, I guess that means the next step is to remove the new connectors from their slot covers, remove the existing back panel connectors from old kerby and replace them with the EPoX back panel connectors and cables. That done, everything connects just fine.
With the I/O connectors on the back panel connected properly to the system board header pins, the next step is to connect the power supply to the system board. This is an old AT-style power supply, with two separate connectors labeled P8 &; P9 rather than the single 20-pin ATX power connector. Actually, the EPoX system board has connectors for both styles of power supply cables, which is a nice touch. I connected P8 and P9 to the system board, making sure to put the black wires on the two connectors towards the center.
With the power supply connected to the system board, its time to connect the header pin wires for the front panel indicators--power, disk activity, speaker, etc. Connecting those went pretty well, with one exception: the Gateway case has a five position connector for Keyboard Lock/Power LED. Position 2 is filled in to prevent you from inserting this jack backwards. Fortunately, the connectors for Keyboard Lock/Power LED were in the correct positions on the EPoX system board to fit the 5-position Gateway connector, but all five positions had pins. So, I gently bent pin 2 backwards on the EPoX system board to allow the connector to fit. All the others went on fine, with no surgery required.
With all I/O and front panel wires connected and verified visually, its time to install the hard disk and connect the floppy and IDE cables. The floppy cable was long enough, and connected without problems, although the EPoX board depends on a connector key for correct orientation. I couldn't find a Pin 1 indicator anywhere, and the original cables aren't keyed. I probably should have just swapped out the original floppy and IDE cables for the keyed cables that EPoX supplied with the system board, but it was easier just to compare them to determine how the red stripe on the cable corresponded to the connector key.
The IDE cable had the same connector key/pin number 1 problem, but I compared it as well. The existing IDE cable is connected to the CD-ROM drive, and required a real stretch to reach the secondary IDE connector on the system board. It worked, however, and I didn't have to put enough stress on it to cause any problems. I used the IDE cable that came with system board to connect the new Seagate 4.3 GB IDE.
The Smoke Test
Finally, I installed the Matrox Millennium PCI video board (which fit fine), carried the box into my office and connected everything up. The system started the boot sequence normally (no smoke), but of course had no operating system installed.
Getting to the CD-ROM Drive
I stuck in the Windows 95 boot floppy that Steve Tucker and I made last night. In addition to the Win95 boot files, it had generic ATAPI IDE CD-ROM drivers that I was hoping would allow me to access the Windows 98 CD. I used the boot disk to create a 2 GB partition on the new Seagate 4.3 GB IDE drive, and then did a format c: /s. DOS promptly informed me that there was insufficient memory to transfer the system files. I backed up, and just formatted C: normally, without making it bootable.
Then, I tried to access the Windows 98 CD-ROM, and that's when the trouble started. When I tried to log to drive D:, the system returned an error message telling me that the CD was "not High Sierra or ISO-9660 format." That's odd, but with Microsoft anything is possible. I tried a couple of other known-good CDs, and was never able to log to the drive.
Okay, obviously the problem is that the system can't see the drive, which means I need the appropriate CD-ROM drivers. I'm sure that somewhere around this place is the original driver disk for the CD-ROM drive that Gateway sent me when I bought the system, but finding it would be next to impossible. No problem, I figured. I'll just bop over to the Gateway web site and download the driver I need. The Award BIOS on the EPoX system board was kind enough to tell me that I had a Mitsumi FX400 drive, so finding the right driver shouldn't be too hard, or so I thought. I'd forgotten how truly crappy Gateway support has become, both on the phone and on their web site.
If this was a Dell box, I'm sure I'd have been able to find the correct driver in 30 seconds flat. Gateway, on the other hand, apparently doesn't think that they should make drivers for old hardware available for download. I searched for "mitsumi" and found a few hits, but nothing for any drive older than a 12x version. I'm pretty sure this Mitsumi is a 4x version. I didn't waste much time on the Gateway web site because it quickly became clear that there wasn't likely to be anything helpful there. I don't know why I even bother to try it any more. There's never anything there that's helpful.
At any rate, the next step was to hit the Mitsumi site. I quickly located the necessary drivers, downloaded and extracted them, copied them to the boot floppy, and made the appropriate changes in config.sys. When I booted with this new floppy, the CD-ROM drive showed up as D: and there sat those gorgeous Windows 98 installation files.
Installing Windows 98
The rest of the story is anti-climactic. I ran Windows 98 setup, answered a few questions, got to the part where it starts to copy the files, and took the dogs for a walk. When I got back, Windows 98 was installed and running. It had detected all my hardware properly, configured itself, and was ready to use. I've heard that a lot of people have had problems installing Windows 98. You couldn't prove it by me. I've installed it only three times now, but on three pretty diverse systems, and it's gone flawlessly each time.
Old kerby still needs some work done on it. I have to add the other 64 MB DIMM, install the Creative AWE-64 sound card, and put in a network card of some sort. Once I have things stable, I'll swap hard drives around, putting the drive in Barbara's current system into this new box (converting the new box to thoth in the process) and moving this new hard drive into the old thoth box. It's a Pentium/133 with 64 MB, which should be plenty to run Windows 98. I also have 2+ GB of unused disk space on this drive, so I'll probably install Windows NT Workstation on it, and convert it to a dual-boot client machine. I purposely avoided running the FAT-32 conversion utility for just that reasonI want NT to be able to see the volume that Windows 98 is currently installed on.
More later on the continuing saga of machine upgrades...
Copyright © 1998 by Robert Bruce Thompson. All Rights Reserved.