How experience with my first PC led me to build my own ever since.

Unlike many of my colleagues, the enthusiasm for the early versions of the PC, the Amiga, the Atari and so on, passed me by. It was partly because I had enough exposure at work to a very hands-on and low-level programming environment that I didn't feel the need to replicate the experience at home, but it was also to some extent because I didn't see much of use being done with the ones my friends had. It seemed a rather large investment just to play those amusing two-person shooting games with a cactus in the middle. However, by the early 1990s the PC magazines were becoming fatter and glossier, and the products on their pages seemed to offer more every year. It was only a matter of time.

The choice of manufacturers was already extensive when the search began in earnest in 1994. At that time, Intel's 486 chip was reaching the peak of its development, but newer architectures were in the works. Dell and Gateway were already prominent alongside IBM, and there were hosts of minor players who sometimes achieved the top spot in reviews only to disappear months later. That observation was enough to make me stick with the biggest players when choosing. What finally settled the issue was the fact that Gateway had opened European production facilities in Ireland, promising faster delivery times and payment in a currency I already had. I still have the invoice for that first PC. It cost the by today's standards hair-raising sum of £3,200. And that was with Windows For Workgroups 3.11, an operating system which looks laughably primitive now. Gateway had quoted a week from the order date to shipment, but they exceeded their target handsomely, so that when I rang back to choose the free application, it had already left the factory, necessitating the return of the default Word 6.0 to be replaced by Excel 5.0.

The initial experience was enjoyable enough. So much software and so much to learn! It soon became obvious, though, that all was not well. The machine tended to freeze at random. It was always a hard freeze, necessitating use of the Big Red Button. Sometimes the monitor's display was bisected diagonally, with only half showing, but mostly it was just a plain freeze. After a while I called the Gateway help desk. They came up with the usual proforma suggestions, whose vacuity was only evident later. Reboot the system, look at various settings, reinstall Windows. Eh… and lose all my carefully customised settings? Thanks!

As time went on, the problem seemed to become worse. The freezes happened more often, and now I observed that if I didn't power the PC off soon after they happened, it couldn't be restarted. The BIOS settings in the CMOS memory were getting lost. Here was where Gateway had done one thing right. With the PC had come a printout of the BIOS settings. I soon became an expert at quickly inputting them again. This was when it became clear that software wasn't the problem. So began the resort to opening the case and removing and reinserting components. This seemed to buy me some time before the problem returned, but it always did. In my efforts to cope, I became expert at removing the motherboard and dismantling the machine entirely, then rebuilding it exactly as I found it.

After several years of this, I finally decided that I had to get another PC. Clearly, Gateway wasn't going to be considered again, so the options facing me were:

Once again, the PC magazines piled up as I checked out the major brands' offerings. They were already far cheaper and better than what I had, but one question bothered me. I was by then accomplished at fiddling with the internals of my machine, and I wanted to know that I could make whatever modifications I wanted later, using generic components as required. There didn't seem to be much information about this topic around, but the issue was settled when I overheard a colleague asking just this question about some component he wanted to put into his Dell, I think it was, and getting a negative answer from the help desk. The case configuration or interfaces wouldn't allow it. I decided that that wasn't acceptable.

The second option very nearly happened. After yet more time at the magazine racks and a few phone calls, I went into a local PC store, expecting a less sophisticated product, but somewhat closer to what a home-built machine would be like, and therefore with fewer compatibility problems. I was resigned to getting my OS in German, a language I could cope with, but didn't want. Before me at the counter was someone who had bought one of the shop's machines which had failed. He was returning it for repair under guarantee, insistent that he needed a replacement, anything at all would do, as it was for work purposes. The clerk was just as insistent that they didn't provide that kind of service. The exchanges went on, the customer explaining why he felt entitled to support and the shop explaining that he wasn't going to get it. A few minutes of this were enough to tip the already delicate balance. Before any other salesperson noticed me, I slipped quietly out of the store again and forgot about them.

It looked like the only satisfactory possibility was the last one. I happened to mention it to a Swiss colleague, whose response was positively offhand. You didn't need much more than a Phillips screwdriver and a place to work. Most cable connections were keyed, so that you couldn't damage things by plugging them in the wrong way, and those that weren't, like the wires to the case, didn't matter. If they didn't work, you just had to turn the connector the other way around.

Suitably encouraged, I went out and bought everything required. The case was the hardest thing to choose, since I had no experience of what mattered, and I ended up with something unnecessarily expensive, but the rest presented no dif ficulties. That first build had only one problem, namely that you have to press down frighteningly hard on those memory sticks to seat them on the motherboard. Once the colleague had shown me how, the machine worked perfectly. As a bonus, I had avoided the "Microsoft tax".

Even with the replacement working fine, and the Gateway essentially redundant, there was still this nagging curiosity about why it didn't work. So, with it not being needed on a regular basis any more, I gave it to Francis, a colleague at work whose involvement with PCs dwarfed mine, to take it apart and see if he could spot the fault. A week or so later he brought it back, saying it seemed to be working OK, although he hadn't found anything specific wrong during his manipulations.

Sure enough, when I plugged it in at home, it was working. It ran for how many days I forget now, but enough for it to be quite clear that the old fault was gone. That was pleasing enough, but it was annoying that I would never know what had been the cause of all that grief. I opened the case once more to see what Francis had done. Nothing major, no new components or anything. But I noticed he had put the graphics card back into a different PCI slot, so out of sheer habit I moved it back where I had always kept it, closed the case for the last time and resolved to forget the problem for good.

…and the fault was back!

It wasn't any of the major components, nor a problem of overheating, nor was it software. The PCI slot I had been using for the graphics card had some kind of sporadic short circuit. As a beginner PC user, I had had no hope of guessing this, and my determination to put things back exactly as I found them whenever I worked on the machine meant that I never found it by accident. For this was Gateway's reputation ruined, and PCs given a bad name for years.

It's an ill wind, though. That bad slot meant that I was forced to find out that I could build my own PC from components. Since that first build worked I've never looked back. I've built three or four machines since then without a hitch, and have been cured for good of buying preassembled PCs (and paying Microsoft for something I don't want).

As for the Gateway P-66, it soldiered on for years. The original Western Digital disk died in 2007, and the monitor is a distant memory, like the ancient dial-up modem and the 2X CD reader. In its last guise, it had two hard drives in cartridges, swappable into the connected slot as required. One, with a whole 240 MB (yes, MB) had a bare-bones installation of Windows 98 SE and a single application to use my old Agfa scanner. I installed OpenBSD 3.5 on the other. Why? No reason, really. Just because I could. It was an exercise in black magic to get it to install at all, and I never really found a use for it. Anyway, the Gateway used too much power and made far too much noise for it to be left on much. It lived out its days under a desk as a reminder of early optimism and dashed expectations.

As time passed, it was switched on less and less. Then, one day in October 2012, after an interval of something like four months, I pressed the power button to find that it had passed away in its sleep. The POST checks ran, but then the BIOS reported that the CMOS battery state was low, and it couldn't read the hard drive parameters. A quick look at Gateway's manual was reassuring. The "System Board" section clearly showed where the battery was, and the heading "Replacing the CMOS RAM Battery" informed me that...

          The clock has a replaceable battery and oscillator inte-
          grated into the device, and installed in a socket on the
          system board. The expected life of this component is ten
          years. It is next to the power connectors, and is easily
          removable by hand.

It had lasted not ten, but eighteen years... good! The sealed unit was, however, soldered directly onto the motherboard... doubleplus ungood!

No conceivable application could possibly have justified the effort needed to resuscitate it, and the next day it went to the Great Server Farm in the Sky.