<In Which I Build A Computer (Spring of 2001)

My old computer was a dual Pentium Pro with its chips running at 200 MHz. There were 250 MB of RAM on board, and two 8 GB IBM hard drives running off an Adaptec 2940 SCSI board. The Drive C (boot drive) was for the system and applications, and Drive D was for data. Drive D was nearly full. The graphics card was an old AccelGraphics AccelPro with 8 MB of RAM.

The Myst III demo laughed in derision at my graphics card.

The system included a vanilla CD-ROM, floppy disk, Iomega 100 MB Zip Drive, and a Colorado T4000S tape drive.

I was running Solidworks, AutoCAD, 3-D Studio, Accurender, and the usual Microsoft Office 97 tools. The operating system was Windows NT with Service Pack 6.

The software and my data were outgrowing my machine. The time had come to get a new system. I decided that Window 2000 would be the best bet for a new operating system, and started thinking about hardware.

Selecting the Processors: Do I Let Some Blue Guys Tell Me How It Is?

The first big decision was what to use for a processor. After researching this subject and compatibility issues, I decided to play it safe and stick to Intel. This “play it safe” philosophy, usually reflected by my selection of major brand names, was a big influence for most of my buying decisions. This, combined with a desire to build a system with more power and margin than I probably needed almost certainly resulted in my spending much more money than a more skilled and experienced computer person might have had to.

I narrowed the processor decision down almost immediately to either the shiny new Pentium 4 or a pair of Pentium 3 processors. While the Pentium 4 was impressive, there were issues on socket and software compatibility, and the fact that software had to be specifically engineered to use its fastest features. How long before my software caught up? In addition, much of my current software was specifically written to take advantage of dual processors. In spite of the cool Blue Man Group advertising for Pentium 4, I went for the Pentium 3 dual-processor configuration. Hey, a few months back these blue guys thought Pentium 3 was the way to go.

This also demonstrated a second buying rule-of-thumb that has also served me well in computers: find the “cutting edge technology” component, and step down to the next one in line. I sometimes have no real rational reason for doing this, but believe I have had less trouble with software support, driver availability, and general gremlin problems. It is probably related to my reluctance to buy Version 1.0 of anything. This “not-quite-cutting-edge” concept has some exceptions, of course. For example, I try to keep up with the latest updates for virus software, drivers, and existing operating systems.

Off-The-Shelf vs. Custom Built: Taking the Plunge

I checked the usual vendors for in-the-box systems: Dell, Compaq, Gateway, and a few others. All of these had attractive and easy-to-use customizing pages on their web sites. Unfortunately, all offered only a limited selection of components, many with uncertain pedigrees. Although all of them had excellent choices, even in dual-processor, none of them offered exactly the system I was looking for. I began to think about building a system from scratch, which I had not done for years. This was not an easy decision, given the siren song of support, warranties, and turn-key operation offered by these in-the-box vendors.

I am not a computer wizard, but I am not entirely dim and have spent a lot of time “under the hood” of my current systems, upgrading them and messing with their little silicon heads. I felt like I could do this. I decided to give it a shot. God help me.

Throughout the entire process that follows I relied heavily on advice I found on the Internet. Jerry Pournelle and Robert Thompson gave me invaluable information based on experience and knowledge far beyond mine, and many of my buying decisions reflect stuff I read on their sites. There were other useful sites of general computer information including PCGuide.com, Arstechnica.com, Anandtech.com, and many others. Also highly valuable were the individual sites for each component vendor. Most of these had very detailed information, often including installation and user manuals online. I don’t know how the hell I ever did anything before the Internet.

Motherboard: Intel Inside

Obviously, step one was finding a motherboard for my two processors. Again, drawing on others’ experience and my own gut instinct to “play it safe,” I decided to stick to Intel. Checking with Intel, I found only one board that stood out for dual processors: the OR840 (another option was the MS440GX, which ran the Pentium 3 Xeon. I am embarrassed to admit this, but after all my research I am still not absolutely clear on exactly how the “Xeon” differs from the vanilla Pentium 3).

I downloaded the massive Acrobat documentation on the OR840 board, and found that it had what I needed: support for 2 GB of RAM, an AGP graphics slot, and a new and updated 840 chip set. Also included was an ATA/66 IDE connection, and a built-in network port. The version I got did not have on-board audio, which would have been nice. Maybe a 64-bit PCI setup would have been nice too, but although I had a few components that could have used it, I found no components that demanded it. The board had the SECC2 interface for the processors. Again, maybe not the latest and greatest, but fine with me.

The board was hard to find. I finally ordered a boxed board from LEM Computers, which delivered it with no problems. Included with the board were some cables, a short manual, processor retainers, and other goodies.

Purchasing Parts: Fry’s Electronics

I bought a lot of my components at Fry’s Electronics. There are actually two stores within easy driving distance from my house. I live in the heart of Silicon Valley, about two blocks from the Netscape complex in Mountain View (fortunately for many dotcommers the typical tilt-up building is too short to commit suicide by jumping off the roof, but there are a lot more empty buildings around here than there used to be).

When I moved out here in 1981, Fry’s was relatively new, and was geek heaven. I was fascinated by the combination of state-of-the-art computer components, girlie magazines, and junk food that could be found nowhere else at that time. The first store I went to down near Lawrence Expressway was tricked out to look like the inside of a computer, with giant chips, capacitors, resistors, and wires scattered throughout the store. The newer store in Palo Alto has a Western theme, with mining cars and plastic full-sized cowboys and horses all over the place. The one in San Jose is apparently Egyptian, like the Luxor in Vegas, but I never go there. The stores are highly successful, and new ones are always being built. They open early in the morning, and are open late in the evening. They are a lot of fun to walk around in.

There are a few simple rules for shopping at Fry’s. These are some of mine, and I’m sure you have a few more of your own.

1) If you don’t know pretty much exactly what you came in for, you are probably in trouble. There are knowledgeable people there, but they are not easy to find, and there is no way to tell them from the people who are simply talking through their hats.

2) Many packages on the shelf have a little white sticker on them showing that they have been returned by customers and repackaged. The sticker assures the buyer that all warranties will be honored, and that this is a perfectly good product. Don’t touch these with a ten-foot pole. Warranties aside, very few customers seem to be conscientious enough to replace every screw, paper, clip, cable, or plug back into the box before they return it. Often, something important is missing. Search the shelf for unopened boxes—they are usually there. I have learned this one the hard way.

3) Try not to walk out of there with anything you didn’t come in for. Like shopping at Costco, this is not easy.

Processors

The processors were available at Fry’s. I used 933 MHz versions, partially because the OR840 board had a few issues with the 1 GHz versions (apparently fixed in more recent motherboards), and partially in deference to my “not-quite-cutting-edge” rule, which often also gives me better prices (there is nothing like a new piece of hardware coming out to drop the prices on the slightly older stuff.) I also bought the cooling fans for the processors, selecting big, clunky heatsinks with dual ball-bearing fans.

Here is where I hit my first small crisis of doubt. The hookup of the heat sink to the processor seemed obvious (no instructions at all came with the heat sink), but I was stricken by paranoia, wondering if there was something crucial about this that everyone knew except me. After checking with a few web sites, and finding nothing specific, I went ahead. I spread a thin layer of conductive grease on the metal square that seemed to be the business end of the processor, and clipped the heat sink to it. I repeated this for the other processor, inserted the retainers into the motherboard, and plugged them in. There were two fan connectors on the board, and four fan plugs, but the heatsinks came with a splitter cable that fixed this nicely.

RAM

The OR840 board needed 600-800 MHz Rambus memory, and supported up to 2 GB. I’d been there before—2 GB assumes you are using 512 MB chips, which are hideously expensive and hard to find. Two 256 MB chips would be just fine for now, thanks, and I could upgrade to 1 GB someday if I needed it. I selected Kingston for the brand, based on advice from my sources, and went whole hog for the 800 MHz ECC. I had no real idea if this would gain me anything tangible over 600 MHz RAM (and still don’t), but what the hell. Not too surprisingly, Fry’s didn’t have these chips. I went online, then, to Outpost.com, which up until now has been a reliable source for computer stuff for me. I ordered two chips, and started waiting.

Hard Drives

Again I chose to go with SCSI, not because it was that much faster than the modern IDE connection, but because I have a lot of legacy hardware laying around that I may want to hook up later, and I just plain like working with SCSI. I have had real good luck with my old IBM 8 GB Ultrastar drives, so I decided to go with IBM again. This in spite of the fact that they are almost as noisy as my old industrial drive was. A look at the IBM site showed me a set of Ultrastar 36LZX Ultra 160 10,000 RPM drives that filled the bill (there were brand, new 15,000 RPM drives, but the “not-quite-cutting-edge” rule said this was overkill). I needed a little more elbowroom on my Drive C, so I ordered an 18.3 GB for that. I needed a lot more elbowroom on my Drive D, and picked a 36.7 GB drive for that. There were larger drives, but this fit in well with my major backup routine, which involves backing up all my data to my Thinkpad 770 laptop’s Drive D. Unfortunately, the current Thinkpad Drive D held only 8 GB. I would have to replace it. I found an IBM Thinkpad Travelstar hard drive that would hold 32 GB, so this would work out close enough.

I found the Ultrastar drives at Fry’s, filled out the order, and went to the checkout desk to pick out the hardware. They supposedly had four of each in stock, but could find none! They were suitably chagrined and apologetic, but I had no drives. That is one heck of an inventory discrepancy. I think they need to check on whether or not some of their inventory is “walking” out the door.

So, I ordered the two Ultrastar drives from Outpost.com, as well as the Travelstar. The 36.7 GB drive arrived immediately, while the 18.3 GB drive took a bit longer. When it arrived, it was not exactly the right drive. It did not have the 68-pin connector, but rather an odd 80-pin connector that I had never seen before. There was no power plug at all. This was a new one on me, but I wasn’t going to deal with it anyway, so I didn’t worry about it. A quick exchange with Outpost.com, and I had the proper 68-pin, 18.3 GB drive in my hands.

The Travelstar drive was apparently backordered. Unfortunately, I had already moved the Thinkpad 8 GB drive (my old Drive D) into the Drive C location, and set it up as my new, roomier Drive C (replacing the 5 GB drive that had come with the laptop). I had already enthusiastically loaded up a brand new Windows 98 OS, replacing the old Win 95, and reinstalled much of the other software. This now left me without my primary backup routine until the new Thinkpad Drive D came. Dumb. I should have waited.

SCSI Card

Picking the brand of this was easy. I’ve never used anything other than Adaptec. I spent a lot of time trying to pick from three cards: the 29160, 29160N, and the 39160 2-channel. All of these cards can handle legacy SCSI devices, but warn that hooking up a slow device to the Ultra 160 SCSI ports will slow all of the ports down to the legacy speed. I wasn’t sure that I wouldn’t need a 2-channel device to solve this, but after some checking found that the extra 50 and 68-pin legacy ports on the 29160 would work just fine without slowing down the fast ports. I settled on the 29160 card, and picked one up at Fry’s.

Sound Card

This went quickly. I figured a Soundblaster brand would be a safe bet, and just picked one off the Fry’s shelf. I got a Soundblaster MP3+5.1, which was probably a little overkill, but I plan on doing multimedia work eventually, and went with a little more than the minimum.

Graphics Card

This was kind of a bear. I needed heavy graphics power, the technology changes almost moment by moment, and I was looking for something that would work for video and 3-D as well as 2-D graphics. I spent many days browsing web sites, reading reviews, and checking out chips. Unfortunately most of the graphics card reviews were based on how well they worked for games, which, while nice, was not my strongest criterion. The NVIDIA GeForce chip seemed to be a good all-around performer, and based on what I read, I finally selected the ELSA Gladiac Ultra, with a GeForce 2 chip. The new ELSA Gladiac 920 with the new GeForce 3 chip was tempting, but its newness, and my “not-quite-cutting-edge” rule kept me honest. Fry’s did not usually have a lot of Gladiac Ultras in stock (maybe a good sign?) but I got lucky and found one on the shelf. The Gladiac Ultra design had an add-on video plug-in board option that, while I didn’t need to buy one now, may be useful farther down the road.

CD-ROM

I decided to go for a CD-RW device that would give me the ability to burn CDs for music or, more importantly, data. There were many good brands, but a review in one of Pournelle’s BYTE columns decided me on the Plextor Plexwriter 16/10/40A. I ordered one from Outpost.com. The CD-ROM was of the E-IDE ATAPI type. Outpost.com had a deal which included a bunch of blank CDs for no extra cost, so that worked out okay, too.

I briefly toyed with getting a DVD device, but I don’t use my computer for watching movies, and the technology for writing DVDs was changing rapidly, with radically new products just over the horizon. I was confused by the many standards, and would really want to do only two things with it: save large amounts of data, and write video that could be read by a commercial DVD player (to transfer home movies). This latter ability is quite new, and quite expensive, with Pioneer and others starting to come out with some good stuff. I chose to wait a while on this.

Removable Disks

The 3.5 floppy was easy. I picked it off the shelf at Fry’s. I got a TEAC drive in a box for a little extra money rather than using one of the shrink-wrapped jobs lying in heaps on the shelf. Maybe some computer wizard could fix a messed-up floppy drive, but I needed something that worked right out of the box. I have heard that many people are having trouble with Windows 2000 and floppies, but fortunately I never ran across this (so far.)

I added a 250 MB Iomega ATAPI ZIP internal drive, also from Fry’s.

Case

At this point, I really needed a case. The PC Power and Cooling Company came highly recommended, and I checked it out. The cases, as described, were high-quality, but only came in two or three basic models. The “Economical” version was too small for what I had in mind, and the “Professional” version was massive overkill. Reluctantly, I looked elsewhere.

I went back to Fry’s and looked at what they had in stock. The nice part about this was the ability to get some hands-on examination of the cases. I was looking for a certain number of bays, at least a 300-watt power supply, and a certain amount of access convenience. The case I use now requires a screwdriver to do anything at all on it.

I liked the way the Antec cases were laid out. There was a lot of room, good construction (with the possible exception of some of the plastic parts), easy access (with an actual latch! Took them a while to think of that!) and a lot of extra fan locations. I also really liked the drive cages that easily snapped out of the case to be worked on. I picked up an Antec SX-1030, and brought it home.

The first immediate problem was that the fans stocked with the case were cheap, and did not have the 3-wire connectors required to plug them into the temperature-control fan plugs on the motherboard. I went back to Fry’s and picked up three 3-wire ball-bearing fans. I snapped them into the neat little purple fan holders and plugged them into the board. Interestingly, one of the fan holders snapped directly inside the front of the hard drive cage. I thought this would be great for keeping my hard drives frosty, and put a fan there after asking around to allay my concern that a fan so close to the drive might interfere electronically with it. There was a fourth fan position in the front of the case, but figured I could just leave that space open for air intake. Besides, I was out of motherboard fan plugs. The front of the case had a purple plastic device with slots in it, apparently to retain the ends of extremely long PCI boards. When I first started working with computers it was quite common for boards to be huge, occupying the whole length of the case. Nowadays even powerful cards are relatively small. Some are simply a connector, and a nearly empty card with a chip stuck in the center. None of the ends of my cards came even close to the other end of the motherboard, which is kind of nice since it leaves my IDE connectors open.

Computer Esthetics

A little aside: I love looking at circuit boards. They have always been interesting, even back when I had my PC-AT, and the boards still had striped cylindrical resistors and those cute little 3-legged transistors standing on them like Martian war machines. The traces were wide enough to drive a car down.

Nowadays the boards are works of art. The first really beautiful board I saw was the single board of the Sparcstation 10 I used at Stanford, a wonder of miniaturization compared to anything else I had seen at that time. When I unpacked my new OR840, I just stared at it for a while. Most of the components were barely visible to the naked eye. I got the creeps thinking how easy it would be to accidentally make a tiny scratch on its surface and shred dozens of hair-thin traces. The 82840 Memory Controller chip had its own little heat sink. The resistors and capacitors were now tiny little rectangles (dang, I used to know how to read those little colored stripes), although a few big capacitor cans still sat here and there on the board. A pedestal near one side, the DC-to-DC converter, had some bigger components, including metal doughnuts wrapped in fine copper wire. In one corner, three iron doughnuts squatted, wrapped by several turns in copper wire thick enough to pull a car. I had no idea what they are for. They seemed out of place among all the microcircuitry.

The other boards are the same, beautiful wonders of miniaturization. The ELSA graphics card had a big, square chip in the center with a heat sink and its own fan. Two other heat sinks on the chip probably concealed the on-board RAM. The heat sinks were anodized a pleasing green color.

Of course, this doesn’t even take into account the invisible parts, the molecule-sized components of the circuits deep inside the black ceramic of the chips. I still have a poster of a huge enlargement of a 486 chip, looking like a great city seen from an altitude. In one corner of the poster, like a tiny period, is one of the actual naked chips. And this is primitive compared to current designs.

I was amazed, and proud of a civilization that could do these things. To me, this was far beyond electronics. This was near-sorcery, and if it all worked when I started it I would be astonished.

Assembly

I started by installing the motherboard in the case. Big case, very easy. The processors were already on the board. I installed the graphics card in its slot, and the sound card and SCSI card in two PCI slots, leaving plenty of cable room between.

The hard drives screwed nicely into their little clip-out cage behind their own little personal muffin fan. I labeled each one, after making sure the jumpers were set to make the 18.3 drive SCSI address #0 and the 36.7 GB drive address #1 (I hate jumpers. I love those little switches that have the SCSI number that changes as you switch it.) I left the rest of the jumpers as is, after consulting the manual.

There was another clip-out cage mounted above the one with the hard drives, sitting behind the 3.5 inch openings in the front of the case. I mounted the floppy and ZIP drives one right above the other. I was counting on myself being bright enough to not be constantly trying to shove the wrong kind of disk in the wrong hole. (A later note: yeah, right. I’ve shoved a floppy in the Zip drive at least three times so far. I was breaking a cardinal rule of design engineering: “always assume the user is an idiot.”)

More Power!

I stood back and looked at the interior of my computer. Many fans. Two processors. Two screaming hard drives. There were massive heat sinks on the graphics card (with its very own chip fan!). Some graphics cards of the Pro50 variety apparently have their very own separate power connector. My motherboard has a plug for one of these.

I began to get paranoid about my 300-watt power supply. I began trying to add up all the power requirements for my components, but it became very complex very quickly, since all of them varied wildly depending on what they were doing at a given moment.

I decided to hit this problem with my usual big, dumb hammer, and ordered a PC Power and Cooling Turbo-Cool 450-watt ATX power supply. I would increase my power to a comfortable margin, and at the same time get the reliability benefit of the most important part of a PC Power and Cooling case.

The power supply arrived, and I unscrewed the old 300-watt model. The case had a thin metal plate with cutouts for the fan and plug to which the power supply was attached. This plate in turn was screwed into an opening in the back of the case. There were no other attachment points for the power supply. When I unscrewed the old supply, I noticed that it had been mounted upside down, so the fan was on the right. If I put the new power supply in right side up, the fan would not match the hole in the thin plate. No problem—unscrew the thin plate from the back of the case and flip it over so that the fan hole is on the left. Unfortunately, this damn plate was not symmetrical—the screw holes no longer matched up. I had two choices: mount the new power supply upside down or fix the plate. I spent an annoying half hour with a hacksaw removing some small chunks of the mounting plate. After that, it went together just fine.

No doubt more experienced people are laughing at me right now, knowing that it makes no difference whether a power supply is mounted upside down or right side up, but I didn’t know that for sure (it makes a difference to some disk drives) and figured I’d do it right and play it safe.

Emergency: The Old System Gets An Organ Transplant

Meanwhile, my old dual 200 MHz Pentium Pro started making groaning noises. This was not the first time. A little examination showed that the fan in the power supply was making noises. Maybe my idea to build a new system had not come a moment too soon. I usually leave my machine on constantly, but I got nervous thinking what could happen if the fan in a power supply seized up, and decided to switch it off overnight. This was not a decision made lightly—several times I had done this and the machine was not real good about switching back on again. I was right to be concerned. The next morning I switched on again, and the power supply whirred and groaned, and nothing started up. I had the case open already. I looked at the two processors—the fans were not turning. I was not getting juice. Quickly I switched off. Had I just delivered off-spec electricity to an entire electronic system? What would that do? I pulled the old AT power supply with its stupid push-button switch. Yipes. 230 Watts. I had not built this system, but inherited it and added a bunch of things. Maybe I had been playing fast and loose with my power requirements? The power plugs into the motherboard were scorched brown and brittle where the red wires were. That couldn’t be good. I was now painfully aware of my primary data Drive D sitting in this possibly burnt-out hulk, and its primary backup, the Thinkpad, with a big empty hole where its data Drive D should be.

I remembered plugging in an old 486 system a few years back that had been sitting on my shelf unused for a while. It had exploded! There had been an actual explosion like someone had set off a cherry bomb inside my case. I almost had a coronary. I looked at the inside of that old power supply and found that a number of the big capacitor cans had literally blown their ends off. Examination of the intact ones showed a precise scoring in the metal of their ends, like a burst disk. They were apparently supposed to do this! Who knew? Power supplies. Eek.

I still needed the Pentium Pro for a while, so I went off to Fry’s and pulled a 300-Watt AT unit off the shelf. I was not too concerned about this being a five-year warranty item, so I just got the first one that looked good. About 30 bucks. I hooked it into the old box, and it started up like a dream. All my data seemed intact. I could breath a little easier again. Hopefully, I had the power I needed now (there have been no startup problems since.)

Waiting for the RAM

Now, there was the wait for the RAM to come in. Another week if I was lucky. The room was full of boxes, cardboard, foam peanuts, and bizarre packing cushions. I would discard no shipping material until I knew that everything worked. This is not only a good idea for returns, but it is well-known that nothing tempts the computer gods to do you a disservice like tossing out your shipping boxes.

* * *

April 13. Every time I called Outpost about the RAM, the delivery date was shunted to a week later. This happened three times. I cancelled the order, and located some at CDW.com, who promised the chips would be in my hand within a week. We would see. Fortunately I had finished my taxes before I started all this.

I also cancelled my Outpost order for the Travelstar 32 GB drive for my Thinkpad, which was also developing delivery-date “drift.” I ordered another one from Solutions 4Sure. They supposedly had some in stock. I didn’t like ordering from so many different outfits, but the stuff I was looking for was hard to find, and when I found it in stock I had to snap it up. It’s probably my own fault for being anal about the brand names. Aside from these items, Outpost had been very good about deliveries, and had been spectacular about handling a return that was probably my own fault. I think the hangup with these backordered components was probably a problem with their own suppliers.

April 14. The Travelstar drive for my laptop arrived overnight! Now that was more like it. I had already pulled the old Drive D from the Thinkpad expansion bay and moved it into the C-Drive slot on the other side of the machine. I removed the drive caddy (the metal enclosure around the drive that adapted it to the Thinkpad hookups) from the old 5 GB drive, and pulled the new drive from its box. Oh, God. It rattled when I turned it over in my hands, like something was loose inside. I went to the IBM website, and fished for the FAQ on these drives. To my relief, one said, “My drive rattles when I move it. Is this normal?” The answer was “yes.” It was apparently part of a new head locking mechanism. No problem. Thank God for the Internet.

I hooked the new drive into the caddy, and plugged it into the expansion bay. The thing was half the size of a pack of cigarettes, and held 32 GB(!) To me, this was actually spooky. My first machine, a PC-AT, had a 20 MB drive I had to carry with two hands and cost four times as much. I started the Thinkpad, and used good old FDISK and Scandisk to prep the drive for use. I then wasted no time in duplicating the data on my old Pentium Pro Drive D. I could breath a lot easier now. Without backup I had felt like someone driving through Death Valley in July without a spare tire. The whole dump occupied a little less than a third of the drive. I wondered how long before 30 GB would start to look crowded.

April 16. CDW delivered just one of the RAM chips, and promised the other soon. Typically for the mail-order computer industry, the chip—a little bigger than a large stick of gum—arrived carefully packed in a box large enough to hold a Volkswagen (don’t computer people have any small boxes?) I am cornering the world market in foam peanuts.

For fun, I compared the little 256 MB RAMBUS chip with my old 4 MB Above Board AT, a full-length monster expansion board from my old AT computer. It had a piggyback board, and both boards were thoroughly coated with chips two levels deep. The RAMBUS chip was also $300 cheaper. Wow.

I make these comparisons between old stuff and new stuff a lot. I suppose more experienced computer wizards have long ago stopped being fascinated by the exponential improvements in computer technology. I never have gotten over this, and still get my mind completely boggled at regular intervals. Part of being a neo, I guess.

Now I had to wait some more. One chip wouldn’t work by itself.

April 18. Took the family to Marine World. I not only Saw the Elephant, I rode on it and busted my rump on its horrible rolling spine. When I got home late that night, the second RAMBUS chip was sitting in its giant cardboard box on my steps. I went out and installed the two RAM chips in sockets 1 and 3 (not exactly an obvious pattern, but that’s what the manual is for.) The motherboard company supplied terminator boards that had to be installed in the empty sockets.

It was too late to do anything, so tomorrow would be the big day.

April 19: Startup

I hooked up an old AT keyboard and the mouse from my Thinkpad, and plugged in an extra monitor. I plugged in the power cord. Nothing exploded--a good sign I would think. I connected the built-in network port to my hub.

Just to keep the variables to a minimum, I left the ZIP drive unhooked, and unplugged the sound card. These would be easy to hook up later if everything else worked.

I hit the front button, and everything fired up. As yet I had nothing in either the CD-ROM or the floppy drive. The fans spun. I checked all the little fans on the various processors and everything seemed to be moving that was supposed to be. My old Dual Pentium Pro machine was purring on the same desk, and like my new machine, had its case open. I noticed that the new machine, in spite of having approximately twice the number of fans, was still somewhat quieter. On the screen the standard startup began, and the SCSI bios kicked in. Immediately I hit the controls to activate the SCSI panel.

My first step was to check out and format my two hard drives. The Adaptec interface was comfortably familiar from my old machine. I ran diagnostics and format procedures on both disks. Meanwhile, I noticed that I had apparently not yet decided whether to hook my IDE cables to the primary or secondary connector, and that the business end of the cable was still hanging loose. Oops. After the SCSI procedures, I exited and switched off to plug the cable into the primary connector.

The OR840 came with a CD-ROM that apparently contained drivers and such. Studying the “readme” file seemed to indicate that most of these were designed to assist with installations of Windows 95 and Windows NT 4.0. I decided to just start right in with the Windows 2000 CD and see if it could field all this stuff by itself.

I stuck the Windows 2000 CD into the CD drive, and fired it up again. There seemed to be no problem with the IDE interface, and the CD drive worked fine. The system moved past its bootup routine, and moved smoothly into Windows setup. I walked through a very familiar setup interface, including partitioning and high-level formatting of the two hard drives. Having installed Windows NT more times than I remember, I thought I could almost do this in my sleep.

A snag in the networking area…I let the network Wizard go by during setup so I could work on the settings later. Maybe not a good idea. The Win 2000 network interface was not quite the same. It took me a while to even locate the control panels for TCP/IP settings, and the network settings had a bunch more variables than I was used to. I fiddled with it to the point where I could at least see my other computers on the network. I installed Netscape, and it flawlessly plugged into my DSL modem and went on line.

With my backup Thinkpad visible to the network, I was able to grab my 8 GB of data from the Thinkpad’s Drive D (where I had backed it up earlier from the Pentium Pro) and transfer it over to my new, 32 GB Drive D. My data was now in three places, not counting the tape drive. Bliss.

I then installed the Zip drive and sound card without any problems.

I was in business! With no small satisfaction that my expensive hardware purchasing decisions were apparently justified, I unplugged the old Pentium Pro engine and set it over in the corner. I moved the new Pentium III into its place, hooked it up to the big monitor, and plugged in the Pentium Pro’s keyboard, Logitech mouse, printer cable, and other items. No sweat. I fired it up. Who’s the bomb?

The Horror, the Horror…

The sin of hubris is a particular favorite of the vengeful computer gods. It ranks way up there, much higher than other sins like throwing away shipping boxes, refusing to back up files, or not using an uninterruptible power supply.

The first thing I noticed when Windows restarted after I moved the machine was the “new hardware” dialog box, which had apparently detected my old Photosmart printer connected to the parallel port. I’d never had Windows see a printer as “new hardware” before. I cancelled this, since I preferred to install from the CD, which always worked best for me. My preliminary desktop appeared. There was something wrong.

No cursor. I moved the mouse around. Nothing. Well, dang. This mouse was a Logitech 3-button mouse. The one I had installed originally was a Mouse Systems 3-button mouse. Could it really be that different? Why hadn’t Windows detected the new mouse and installed drivers? I rebooted. Same thing. Double dang. Using the Byzantine and awkward keyboard commands (that I can never remember) for dialog box selection, I painstakingly got to the Hardware Installation interface, and ordered it to find my mouse. Nothing. I went into the control panel, selected “Mouse,” and checked the hardware tab. Blank. I went to the “System” icon, found the Device Manager (more like the Windows 98 interface than NT’s, thank God.) There wasn’t even a line for mice or other pointing devices. I rebooted again. Nothing.

I am not the most knowledgeable person when it comes to computers, but I have a long history with them, and have learned a few things in the school of hard knocks. One of these is that Windows can get permanently corrupted by an amazing number of things, and despite Rescue Disks, and Backups, and Norton Utilities, and all the wonderful repair tools available, the only thing that really works all the time is to scrape the hard drive down with a format and start from scratch. I did this to my Drive C. I started reinstalling Windows 2000. The cursor was not there, even in the early startup screens, and certainly not in the Windows installation dialog boxes. Little beads of sweat began to appear on my forehead.

I checked the obvious. I reattached the Mouse Systems mouse that had worked before and started the installation again. No cursor.

What had I done? I checked the less obvious. Had something happened to the mouse connector on the board? It was the right connector. I double-checked, and besides, the keyboard was working fine, in spite of being one of my old favorite AT keyboards. Was something in the connector? I blew out the connector port with an air can. Still nothing. Good God, had I busted some connection to the motherboard by plugging in the new mouse too firmly?

At this point I also noticed that since I had reformatted only my Drive C (or Drive 0 to the SCSI manager) during my recovery efforts, that Windows had gone off and assigned “Drive C” to my data drive (which was supposed to be Drive D, and currently full of my data), and then assigned D and E to my CD-ROM and Zip Drive respectively. There may be ways to fix these drive letter assignments in Windows, but I have found that if you change the letter of the CD-ROM drive, the Windows system installer (which is amazingly stupid in this regard) will often continue to look for its files in the old location, and unless you know exactly what files you are looking for and where they are, you can’t get the installer to find them anywhere else. Besides, my operating system was now sitting on my data drive, along with all my data. To get to where I wanted to be I was going to have to reformat the data drive too, start over, and transfer all those files again. Crap.

I did this. Went back to the SCSI interface again, and reformatted both disks again, and started installing Windows 2000 yet again. Most irritating part: repeatedly typing in that damn Win 2000 product key which consists of 25 apparently random letters and numbers. Has anyone figured out how many combinations are possible in this sequence? It might give us an idea of how many Windows 2000 seats Microsoft expects to sell.

Still no cursor. I tried other things. I rebooted and restarted installation from scratch several times, using different responses during Windows installation (dragging my way through the procedures using the bloody keyboard), and my machine could still not see a mouse at all. I did find a way to boot directly from the CD, which allowed me to do a high-level reformat of Drive C without having to reformat Drive D, which saved a lot of time and aggravation.

I concluded that I had somehow damaged my motherboard, and resigned myself to beginning the long process of disassembling my machine. I pulled the little connectors that connected my processor fans to the board, then realized it would be simpler to keep the processors attached for now, and started unscrewing the holddown screw on the graphics card.

Suddenly, from somewhere in the back of my head, I flashed on something I had read on Pournelle’s site. I had switched my machine off a dozen times during this last bout using the front button, the reset, or by normal shutdown procedures, but this was not an AT machine like the Pentium Pro had been, with a hard electrical switch. This was an ATX, with a switch that worked in software. I had read somewhere on Pournelle’s site that there was only one way to REALLY turn off an ATX machine. I shut it down again, and this time grabbed the power plug and pulled it out. No obvious changes. I waited. I plugged in the Mouse Systems mouse. Then I plugged in the power plug again. Although the computer remained silent, the three green lights on my keyboard flashed. Once. Oh ho, I thought. A lesson relearned. There’s “off” and then there’s “really off.”

My machine started up again, and this time when the Windows installation dialog boxes appeared, I had a cursor! The mouse worked! (Later experiments showed me that my machine was allergic to the Logitech mouse for some reason, but at least now I knew how to clear out its stupid little head). I thought how close I had come to breaking down my machine and going to the effort of replacing the motherboard. That was a big one I owed Pournelle and his web site.

I was still not out of the woods, but I could see daylight. I finished loading Windows 2000.

Graphics Startup: I pulled out the CD for the Gladiac board. It was time to get this high-powered graphics card out of 640 x 480 idle and start making it work for a living. I restrained myself from installing all the Wonderful Features and stuck to basic drivers and tools (this is a good rule of thumb: whenever possible on a new system, install only the basic drivers instead of all the great programs included for free on the CD-ROM.) Once the drivers were in, I expanded my screen to a more useful size, and went on from there.

Sound Drivers: Again, I loaded only the simple drivers from the Soundblaster CD-ROM. There were dozens of music and sound programs included, but I would look at them another time.

I now went through the basic housekeeping I always did with a new NT-type system. Creating a user file with my name. Setting up a big paging file. Loading Zone Alarm (Good news: a simple check mark in the network controls—“Disable NetBIOS Over TCP/IP” accomplished the recommended “Shields Up” closing of my network to the internet that had required many binding changes in Windows 98.) Going through some Windows options to get rid of the stupid “web-like” windows and put it back to the old interface that I like to use, and configuring a few other things to be the way I wanted them.

I loaded Netscape (not version 6), Office 2000 and Norton Systemworks. I have found that this is a good acid test of an installation. If it still works after all this, you are probably okay.

It worked, but developed a “tic” where the startup process had to repeat a couple of times before the system kicked in. It would get to the point where the desktop should appear a few moments after the screen goes blank, but instead it would go back to the beginning and reboot. Fortunately after doing this twice it would go to the normal desktop. Then it stopped doing it at all.

I went back, cleared the drive, and started the whole installation over. Why? Because I’m paranoid, that’s why. There were no startup problems this time, and that was the last complete installation I did (so far, of course.)

After Netscape, Office, and Systemworks were in (again) I went in and finished configuring them to my taste. I copied all my mail files and such from the Netscape installation in my other machine, and ran the disk checking utilities in Norton, after switching off most of the automatic junk in it I don't use (at least they don't have "Crashguard" any more.)

For the first time, WinDoctor, a registry analysis device, is available for an NT-type system (it has long been on my Win95/98 system). I ran it after installing the system and the aforementioned software, and it found over 50 errors in the registry, most of which were incomprehensible to me (a long string of numbers which is supposed to refer to some mysterious non-existent .dll file and other cryptic problems). It offered to correct these, which I usually do on a "step-by-step" basis instead of letting Norton decide the repairs. What to do? I was having no real problems (so far). I finally corrected the “errors,” but kept the Undo file, instructing it not to erase itself in seven days. Just in case.

I have a suspicion--no basis in fact--that operating systems develop somewhat organically, like genetic codes, and that software developers leave "junk" in it that doesn't accomplish anything but does no harm, like leg bones on a whale or junk DNA. This is easier for them than starting from scratch, and in an age of gigabyte hard drives, who cares if the resulting OS is a little bloated?

Normally, only computer wizards would ever see this junk, but stuff like WinDoctor drags it out into the light, and I get fifty registry errors on a clean install. WinDoctor has only been available on NT for a few months, so hardly anyone has seen this stuff. Maybe this will embarrass Microsoft into going into Windows 2000 and sweeping up a little.

April 28. Present Day. I went to Fry’s to get a couple of power strips. I looked at GoBack software, which I had read a favorable review of the day before. Thought about it. Oh, heck, I had already installed a bunch of stuff, and I wasn’t having any real problems. I skipped it.

If you’re going to spit in the eye of the computer gods, you might as well hawk up a really big one.

I wanted to see if I could get my old scanner to work with Windows 2000. I was tired of running it off my Thinkpad dock. Like most vendors, Umax cheerfully abandoned their old hardware and didn’t bother to make new drivers for it. They had an old NT driver, I thought I’d give it a shot. After all, I had gotten lucky with an unsupported NT driver for my old Kodak DC120 digital camera that worked with no problems. I installed the driver, and was rewarded with my very first Blue Screen of Death from Windows 2000. A rather pathetic one…two or three lines at the top instead of the lovely screen full of numbers that I had grown to know so well. Well, dang. I booted into Safe Mode, and went to the “Uninstall Program” function. The Umax driver wasn’t there. Double dang. After some failed attempts to use “Last Good Configuration” and a few other tricks, I was forced to go in and pry the Umax system files out one by one, which is kind of like picking dog hair out of a shag carpet, and about as much fun. The system worked again. I believe I may reconsider GoBack.

And here I will leave off, with a lot of software yet to install. Building a computer is a lot like a house of cards, with each step depending on the one before, and a good chance you will end up playing “52-Pickup.”

Still, my machine is damn fast now. I can play the big Quicktime files, and I’m looking forward to trying some rendering. I got the new AutoCAD, but AutoCAD 13 still works. And hey! so does Tahni Deskmate! With a new upgrade! It’s going to be a good day after all.

* * *

A Retrospective… A short look at my old systems.

IBM-PC AT: My first major computer. I purchased this to be able to use AutoCAD effectively in my drafting consulting business. At the time, the PC-AT was the top of the line, and hideously expensive. I added a monochrome display for text, an Aydin 21-inch monitor for color graphics, and a Control Systems Artist I graphics card. At the time, these too were top of the line, and I figured if I was going to spend 8 hours a day staring at it, it better be clear graphics. I ramped up the memory by buying an immense Above Board AT memory expansion board (mentioned earlier) for 4 big megabytes of RAM. God, I’d forgotten all about that extended/expanded memory crap. There was a 20 MB hard drive in the box. Later, I ordered a 42 MB Core hard drive. Again, it was top of the line. I was impressed by its ad campaign, where company officials gleefully dropped a large phone book on a hard drive during an access sequence without ill effect. Also thrown in were a flatbed plotter and other odds and ends.

Not counting the big monitor and the plotter, this monster ran me almost $10,000. The AT alone cost almost $4,000 all by itself. I had to finance it like an automobile, and paid for it for over five years. Cripes.

This was also my first experiment with overclocking. I got a new crystal for the motherboard that ramped it up from 12 MHz to a screaming 16 Mhz. Wow!

The Beast: Some years later. My first scratch-built machine. The 486 chip was now the way to go. I got a Micronics motherboard and away I went. The PC-AT ran AutoCAD okay (if the files were small), along with the other programs I used, but I needed More Power. Fortunately, at this stage of technological development sufficient power did not require Top of the Line anymore. I could find cheap stuff that still worked better than the AT. I could also use a lot of old hardware from the AT in the new machine. I recycled the Core drive as my Drive C, and picked up a surplus Atasi 738 from Weird Stuff Warehouse for my Drive D. This horse was five inches thick, with 8 platters and 15 heads, and held over 300 MB. As I remember, this was too big for the OS, and I had to split it into two logical drives. At the time this drive was build for industrial use, and had a reputation for solid reliability, even though there were smaller drives with similar capacity. It was a SCSI drive, so I had to get my first Adaptec SCSI card to make it work.

I got an Orchid graphics card, and a normal monitor (the stupid Aydin had no regular graphics interface, and was fed through individual red, green, and blue signals.)

The Pentium Falcon: After this I went to work at Rotary Rocket, and got the Dual Pentium Pro machine described at the beginning. When Rotary Rocket folded up, I bought it at a fire sale price, brought it to my home office, and retired the Beast. The Beast sat on my shelf quietly for a long time until I tried to start it to see what would happen and blew up the power supply. It wound up at a computer recycling joint after I stripped the useful stuff off.

And that is where I came in.