Hal is a hardware engineer who sometimes programs. He is the former editor of DTACK Grounded, and can be contacted through the DDJ offices.
Once upon a time, Americans would make major purchases--Conestoga wagons, for instance--using gold coins for payment. Then silver dollars appeared, and gold coins vanished from the marketplace. When I was a youngster, retail purchases were frequently made using silver dollars. But paper money drove them out; it's rare to see a silver dollar change hands these days.
Those of you who had to endure Economics 101 will recognize Gresham's law: Bad money drives out the good. Sir Thomas Gresham, the founder of the English Royal Exchange, first explained this law to Queen Elizabeth I in 1558.
My PCs used to consist of the original, horizontal-format case with a small monitor sitting on top. As I moved to larger monitors, it became evident that the minitower case was better for my purposes. In recent years I've swapped cases on four PCs, once because of a power-supply cooling-fan failure.
Here in Silicon Valley you can buy a PC case (always with power supply) at most any clone outlet. I've always checked Micro Times, a local, free tabloid filled with PC ads, to find the lowest prices. I think my first minitower case set me back $63.00. Most recently, minitower cases have dropped to $39.00.
Except for the plastic front, PC cases are made from cold rolled steel (CRS). Until recently, the insides were all alike: The case's innards had deburred edges, were cadmium-plated (to provide that silvery finish) and were welded to provide a strong, rigid structure.
About 18 months ago, I bought a $43.00 case. It was immediately obvious that the CRS was not cadmium-plated; the sheet metal was an ugly gray. After installing the hard drive, the floppy drive, and the motherboard, I discovered both my hands were bleeding. The individual pieces of CRS had been die-punched, a cheap method of fabrication that leaves the edges ragged because die-punching actually tears the edges of the metal. To save money, the metal had not been deburred. The pieces had, however, been welded together.
Recently I bought a $39.00 case. Once again I found unplated, die-punched CRS. This time the structure had not been welded; instead it had been pop-riveted, a very cheap and not particularly strong way to assemble a frame. I have to admit the structure was strong enough to support a couple of disks, a motherboard, and a few cards--modern computer parts aren't very heavy.
But I didn't want to cut up my hands again. I had an old minitower case whose power supply had failed and which I was planning to throw away on the next city spring-cleaning drive. But the new power supply fit into the old case. So I unscrewed the plastic front panel from the new case and discovered it fit the old case just fine. Now I had a new front panel and a new power supply in a deburred, cad-plated, welded case.
That left me with a useless, pop-riveted steel cage. I noticed the sheet steel looked awfully thin. So I stepped on the cage and it immediately collapsed. I turned it and stepped on it a couple more times and wound up with a much smaller mass of smushed metal, which I tossed in my garbage (no waiting for spring cleaning). Oh, yes: I had to use the new U-shaped cover on my hybrid minitower case because the fit to the front panel changed.
I can now find the occasional ad in Micro Times that offer hand-cutting-free PC cases at considerably more than $39.00. Very few are sold.
I still have the cardboard boxes the various minitower cases came in. All have the same diamond-shaped logo on the side, meaning they came from the same Far East trading company. But the good original cases came from Taiwan and had a net weight of 8 kilograms (KG). The unplated, undeburred welded case also came from Taiwan and netted 7.5 KG. The chintzy pop-riveted version was made in China and weighed 5.65 KG.
If this trend continues, the next version of the minitower case will come from Tibet or Chad, will be made of used aluminum foil, and will weigh 3.25 KG, including front panel, cover, and power supply. But it will cost $29.00 retail, and people will buy it.
The PC-buying public has clearly decided that personal computers are generic products. It matters how much "horsepower" the CPU has, and how much DRAM and hard-disk space the computer has, but there is remarkably little brand recognition and even less brand loyalty.
This is a smart decision by the public because, in fact, there's precious little difference between brands of PCs. Some thoughtful consumers are even making the deliberate choice to purchase systems from low-priced clone shops. Not only do you get a lower price up front, but the PC will have been assembled from generic parts and hence can be easily repaired. A major brand-name supplier like Compaq or IBM may well have used a proprietary motherboard (or such) that can't be repaired three or four years later because the part is out of production.
Indeed, the fastest-growing PC-system supplier is Packard Bell, whose computers closely resemble those sold by the cheap clone outlets. Some reviews of Packard Bell's PCs have noted how flimsy the case seems to be....
The marketing types at the PC-system makers have been flummoxed. The market absolutely rejects any attempt to differentiate PCs on the excellent grounds that anything that's different is likely to be incompatible with some of those warehouses full of shrink-wrapped software. Even refrigerators exist in greater variety than PCs.
So the blue-suede-shoes types wept with joy at the development of multimedia. But it turned out that multimedia meant a CD-ROM drive, sound card, and stereo speakers. After growing 300 percent in 1993 and another 300 percent in 1994, multimedia sales are now growing at 20 percent--exactly as fast as the overall market. Multimedia equals encyclopedias and games, and a PC is a horrendously expensive game machine. And, for that matter, how often do you really need an encyclopedia at home?
According to the business section of my local Silicon Valley newspaper, system makers have decided that future PCs will also be TV sets. I wish them luck. I can buy a 20-inch TV for about $200, which frees up the PC. And The Wall Street Journal has begun to print articles about contention within the home for the keyboard of the lone PC. Tie up a $1200-$2500 multimedia PC just to watch "Jeopardy?" I don't think so.
The point of all this is that excess costs have been thoroughly squeezed out of the PC market. Which means anybody who expects to charge a premium price for a different-color case or unnecessary bells or whistles is doomed to failure.
Have you noticed that Apple Computer has been losing market share lately? Apple is the company that historically wants to charge more for its computers because they run a smaller selection of shrink-wrapped software. Uh, did I get that right?
In case you missed it a while back, AT&T sold UNIX to Novell. Novell first tried to figure out what UNIX was good for, then sold it to the Santa Cruz Operation for a heck of a lot less than it paid AT&T.
You like UNIX? Stick around. In a few years you'll probably be able to afford it--not a UNIX system, but UNIX itself. The whole ball of wax.
Intergraph, for instance, makes turnkey CAD stations, just the sort of thing that Fortune 5000 companies like to buy. Well, Intergraph is making the switch from UNIX (and RISC) to Windows NT (and the CISCy Pentium Pro). That's almost a billion dollars a year fleeing the UNIX/RISC market for a Wintel safe haven. Guess why UNIX-system makers have, like Apple, lost market share to Wintel the past couple of years?
So why did Hewlett-Packard cut a deal with Intel to make its future CPUs compatible with Wintel systems? Why is IBM widely rumored to be developing the Wintel-compatible 615 CPU? Why have both the SPARC and MIPS camps announced that their future CPUs will feature hardware support for x86 emulation?
"The war between RISC and CISC is over and RISC won." I've been reading that for over a decade, and I always get a big kick out of it. You see, I used to write the newsletter DTACK Grounded; then I wrote for the late Programmer's Journal, and more recently for Dr. Dobb's Journal (see the January 1994 DDJ, for instance). This publication trail proves I've consistently asserted that CISC would emerge triumphant over RISC. Did you notice how Intel has prospered these past dozen years, years in which RISC folk incessantly forecast Intel's imminent demise?
I wasn't the lone ranger; a few other sensible techies remained in the CISC camp. But I'd guess that the RISC camp, at its peak, included 98 percent or more of the techie crowd who follow such things.
There may be several reasons CISC won. I can tell you my reason for supporting CISC during the long, cold winter years when I was ridiculed for doing so. CISC code is twice as dense as RISC code. RISCs require twice as much (in bytes) to accomplish a given amount of work. That includes the PowerPC.
Which means RISC needs more hardware: twice as much bus bandwidth and twice as much L1 and L2 instruction cache, to get the same work done. And while RISC system makers could and did provide those additional resources, they made a critical error: They failed to notice that more bandwidth and bigger caches cost more money for the same performance.
The mass personal-computer marketplace will absolutely not tolerate anything that costs extra money without providing a corresponding benefit.
To the extent that RISC prospered, it prospered in the UNIX arena, where high prices prevailed and there were no personal computers. But the PC market, with its economies of scale, is a huge black hole that is sucking in everything nearby. The RISC folk are straining to keep their heads above the event horizon, but they're doomed. You don't think Intergraph is the only company fleeing the high costs of the UNIX/RISC camp, do you?
Once upon a time, a supercomputer would support 2000 scientists. Each worker would submit jobs for batch processing. If the supercomputer wasn't down, the job would be completed and the results printed in about 24 hours. Now, imagine the supercomputer as a large black box. What's the most economical black box that'll do the same job as the supercomputer? Why, a cluster of desktop workstations! Yesterday, the workstations were RISC based, but tomorrow they'll be Pentium Pro based.
Did you notice that bankruptcy liquidators just auctioned off the assets of Seymour Cray's Colorado Springs supercomputer company? And that the major "asset" was a completed, up-and-running Cray-3 supercomputer that nobody wanted to buy?
Another black box: It takes a MIPS-based Silicon Graphics workstation about an hour to produce one computer-generated frame for a movie such as Jurassic Park. (Workstations get faster with time but the amount of computer-generated detail demanded by movie patrons does, too, so that hour is pretty much a constant.)
Movies use 24 frames per second, so it takes 60 days for one workstation running 24 hours a day to generate one minute of a movie. Given a z-buffer, doesn't it make more sense to have a roomful of (a black box full of) Pentium Pro systems (already under $4K/system) than a roomful of more-expensive RISC workstations?
The point is, Wintel computers provide the most performance for the least dollars. Soon, any computing job that can be performed by a Wintel computer will be performed by a Wintel computer. This has everything to do with thin, unplated, pop-riveted, hand-cutting computer cases. All of the excess costs have been squeezed out. Gresham would have instantly understood the situation.