March 2000

Tiny flames licked at the bottom of the new hard drive, as a dark, gray, acrid smoke billowed out the back of the just-purchased business PC. "Turn it off! Turn it off!!" yelled the irate, red-faced client. "What kind of junk are you trying to sell me!!?" he bellowed angrily.

This little episode in the spring of 1998 was just the most spectacular in a long line of hardware failures that I have witnessed as a computer repair specialist. While working at three different companies as well as an independent "on the side" consultant over the last three years, I have noted a number of occasions on which an apparently working computer suddenly went down in flames -- sometimes literally!

Usually the failure was simply a quiet refusal to boot, or a blank screen, or garbled data on the hard drive. The failures have occurred on both brand-new "clone" machines (as in the first example, built the same day as the failure) or more often with Packard-Bell, Compaq, and other name-brand PCs. So while there was a wide range of variation in the hardware, the most common factor among these failed, damaged computers was a piece of software: Windows98.

When I published my Special Report last month on OS/2 Warp and the many reasons why it is a superior platform, I took a lot of heat from some computer users who insisted that "bad hardware can cause software to fail, but software can never damage hardware!" They never provided any scientific theories or rational explanations for this belief; it just seemed to be "obvious" to them. Sort of like the way the earth is flat, the sun moves around the earth, and other so-called "obvious" phenomena that fall down when confronted by the facts.

The fact is that software can indeed harm hardware. What is software, anyway? Just instructions. Commands. Orders for hardware to follow. If I take those same instructions and implement them using a set of silicon logic gates in a computer chip, now they are hardware, right? And if hardware can damage hardware, then it really doesn't matter where the bogus, erroneous instructions came from originally -- a software program, or a computer chip. Software and hardware are simply two different versions of the same set of logical conditions. The only difference between them is a matter of cost, that is to say, some forms of instructions are so complicated that they are cheaper to encode in binary form than to etch into silicon.

Another example of the hardware-software duality is the so-called "Java Virtual Machine" or JVM. A Java JVM is simply an emulation of a CPU interface, or coding to a mythical CPU API. The JVM is designed to present itself as a computer chip to the Java application program, which is why there can also be "Java chips" which implement the JVM in silicon (although nobody has seen a reason to move strongly in that direction, yet). What this all boils down to is that any instruction that can be implemented in hardware can also be implemented in software, which means that any damaging signals or messages can originate in software, not merely in hardware. Or to put it more simply, every physical phenomenon has a mathematical formula behind it.

But rather than rely solely on theory, I have come to the conclusion that Windows98 has the potential to damage hardware through years of careful observation. When I worked at one PC shop, it seemed that an average of two customers per week would enter the shop with a suddenly non-functioning PC. The story was always the same: the computer had Windows95 on it, and the consumer had tried to install the Windows98 so-called "upgrade". The repair was always the same -- replace the motherboard, use the same memory chips, the same RAM chips, and all of the same other components. (Okay, once in a while a VGA card was also fried, but that was the exception.) The PC would then boot up, and the point of re-entry was always the same: the point at which Windows98 had first booted in the machine, before it does all of the settings and options. In other words, the first time the motherboard had been subjected to the Windows98 clock cycles instead of the Windows95 clock cycles. Once I even found BURN MARKS on the CPU socket of the dead motherboard. (The CPU itself worked just fine on another motherboard.)

Do you think it can't happen with most computers? Try this, if you dare: Grab an older VGA monitor and put it on a fast PC with a high-end VGA card. Tell Windows9X to use the highest scan rate available on the VGA card. (And what would you use for this statement? Device driver *SOFTWARE* commands.) After a few hours of this abuse, some monitors will burn up. But the VGA card is usually unharmed; it was functioning within its normal design parameters. A set of software commands that told the hardware to do something extreme. Software damaged hardware (using another piece of hardware as the weapon, of course.)

Of course, this kind of CPU or VGA damage does not occur every time, or with every brand of motherboard or PC. And in the case of monitors, almost any operating system could be configured to do damage. But the motherboard example happened with such stunning consistency and regularity, with a clear pattern and a predictable set of circumstances. In every case, the culprit was Windows98, and the fix was to replace some damaged computer hardware.

Which answers the original customer's question: The kind of junk being sold to him by the store owner was Windows98.

Which is also why people who want to keep their computers for three, four, or more years should consider erasing Windows and upgrading to a stable, non-destructive operating system like OS/2 Warp. And it is also why people who don't "get it" should keep a little extra money saved up for replacement hardware, just in case.

I have seen OS/2 run extremely well on 486 PCs with sufficient RAM. On a Pentium of almost any speed, OS/2 runs smoothly and efficiently, with less wear and tear on the hardware than a Windows partition on the same PC. Elite CPUs like the AMD Athlon also benefit from OS/2 Warp's superior performance and reliability. So there's no need to constantly run out and buy a new computer every two years the way Microsoft wants people to feel compelled to do. (After all, a preload monopoly only works if you keep selling preloads.) Just think of Microsoft products as the inverse of Moore's Law: Windows gets twice as slow every three to five years, eclipsing the gains in hardware speed.

Or just think of Windows as Gresham's Law in digital form: bad software drives out good.

Most recent revision: February 29, 2000
Copyright © 2000, Tom Nadeau
All Rights Reserved.