Reliability Can't Wait: Reflections of a MacBook Guinea Pig

Why are some computers more reliable than others?

This is a question that has bothered me over the years and for which there are very few answers - and considerable speculation.

I'm not a technician or programmer, so here goes my attempt at speculation, which - while not scientific in any way - is the result of over 25 years of using and often fighting with computers.

First off, I've noticed two arguments when it comes to computer reliability. One, which is an entirely logical argument: The more complex a given system, the more likely it is to have reliability issues. The other, also based on logic but also on the variable of quality: The less you spend, the more likely you are to have problems.

...device drivers are the real factor in whether a machine will be rock solid or make a great paperweight.

I'll look at these two arguments separately and then look at specific computers that I've had over the years - and why neither of these arguments, in my opinion, dictates how reliable a given system will be. Rather, the quality and maturity of device drivers are the real factor in whether a machine will be rock solid or make a great paperweight.

Increased Complexity

As computers incorporate additional features, there are simply more things to go wrong with them. Back in the early days of laptops, I kept having modems fail and trackball rollers gum-up in my PowerBook 145B. This was an issue of complexity. An internal modem on a laptop was amazing technology when the PowerBook 145 was introduced in 1991, and even though my 145B was made in 1993, it used the same troublesome 2400 baud modem that early users complained about.

The trackball was simply a mechanical device that accumulated dirt and required frequent cleaning, but to someone used to DOS desktops which had no mice, a stuttering trackball was often mistaken for a hardware problem instead of a simple maintenance issue.

Computers have grown far more complex, with DVD writers, wireless network cards, Bluetooth, and gigabytes of memory. As that complexity has increased, so has the failure rate.

In 1993, it was almost unheard of to have a bad pixel on a laptop screen, but with today's high resolution panels it is considered lucky to not have at least one, with most manufacturers having "pixel policies" in place that determine how many bad pixels make for a "defective" screen. Sadly, this number is almost always higher than 0.

Decreased Cost

The second factor is price, and this is a strange thing to look at, as a low price can be achieved three ways - two of which actually improves reliability.

The first way to lower price, at least with laptops, is to increase bulk. It's a lot less expensive to build an 8 lb. laptop than a 4 lb. laptop with everything else being equal. In an 8 lb. casing everything can simply be mounted wherever the manufacturer wishes, while the sheer amount of case material, however cheap the plastic, will ensure a reasonably rigid and sturdy enclosure for those components.

...updated, mature drivers . . . are the biggest reliability factor.

The second way to reduce cost is to use older technology. A two-year-old video chipset or processor architecture costs a lot less than the latest and greatest, even at the low-end, and these components, while offering less than cutting-edge performance, have also benefited from bug-fixes and firmware updates during their production runs, not to mention updated, mature drivers, which I believe are the biggest reliability factor.

The last way to reduce cost is to specify low-end components that aren't old leftovers, but rather new items built to a price. Here the drivers are just as new as on the high-end models, and, since the components are fitted to mostly low-end models, they rarely see the driver and production updates that once high-end (but now outdated) components do.

It's the Drivers

This finally brings me to drivers and why I believe that the software drivers and firmware written for the many components in our computers are the key factor in determining how reliable a given computer will be.

Let's look back at a few of the computers that I've owned over the years and see how drivers directly affected their reliability, both when new and as revised drivers were written and made available.

PowerBook 5300c: Improved with Age

In 1996 I bought an Apple PowerBook 5300c. This machine came with the much-maligned System 7.5.2 and was probably the most unstable, buggy computer I had ever used. It was so bad that it locked up the very first time I powered it on, not even making it through the process of loading its factory installed OS and extensions to reach the Macintosh desktop.

Fortunately for me, by the time I bought my PowerBook, Apple had already issued the 7.5.2 "PowerBook Update", which was for the most part of prerelease of the much-improved System 7.5.3.

All the instability of that PowerBook could be traced to a single item in the System Folder called "PowerBook 5300 enabler", which was the classic Mac OS equivalent of a device driver to allow the computer's logic board chipset to work on an OS introduced before the computer was. The system enabler was so buggy and poorly written (they had a lot of pressure to get things to market quickly) that most of the 5300's bad reputation (everything except for power connectors and case plastics) could be traced to that humble PowerBook 5300 enabler.

With the PowerBook update and later versions of the Mac OS (8.1 was probably the best on the 5300), that "extremely buggy" computer was transformed from one of the least stable, buggiest, and absolute worst computers I've ever had the misfortune to use into one of the very, very best. With OS 8.1 that PowerBook would alternate between heavy use and its very fast and reliable sleep mode for days or even weeks without a restart. The hardware didn't change, the device drivers did.

Toshiba Portege 4000: Great until Windows XP

In the beginning of 2002, I bought a Toshiba laptop, a Portege 4000, which was a high-end ultraportable model roughly equivalent to the 12" PowerBook in size and weight, though considerably slower, due to its older, low-voltage components.

This machine was not designed to be cheap, but to be small and light. It was meant to run forever on its batteries.

Part of that design was an early version of what is now known as "vampire video", where a portion of the system RAM is used by the graphics chipset. This was before the days of Intel's built-in video, and it used a chipset known as the ALI Cyber Aladdin (made by Acer), with the graphics portion provided by Trident, a low-end laptop graphics chipmaker that has since been bought-out, folded, or otherwise no longer exists. "Vampire video" was generally offered on budget machines to cut costs - and on road-warrior machines to reduce power consumption.

The reason I mention this machine is because the drivers for that chipset vary wildly in quality. The computer came preloaded with Windows 2000 and a whole slew of Toshiba-provided drivers for every aspect of that chipset, including modules for sound, AGP (to enable the graphics), the graphics themselves, and power management. It worked well, was extremely stable, and, except for the poor graphics performance, the system was actually very fast for its time and intended use. Even the lousy graphics performance was adequate for a machine aimed at business use. 2D application performance was stellar, and DVD movie playback was also extremely good - as good as on current laptops when using older decoder software.

Where things went bad was when I installed Windows XP on that machine. Yes, there was a whole slew of Toshiba-provided XP drivers available, and they've even been updated a few times over the intervening years, but in XP there is a sluggishness in the system's power management that just wasn't there in Windows 2000. To make matters worse, the graphics driver produces washed-out color in XP, instead of the beautiful and vibrant color it shows in Windows 2000.

The hardware is exactly the same; only the drivers are different.

The hardware is exactly the same; only the drivers are different. Better color on the built-in display under Windows XP is as easy as uninstalling the graphics adapter driver, but then you also lose power management and most other modern chipset-sourced features. I own three Portege 4000s and like them a great deal, but I won't run Windows XP on them because of the drivers.

The MacBook

My last example is Apple's current MacBook.

I had a MacBook until last month and was blown away by the superior design, construction, and speed of this system - when it worked. I went through three MacBooks and four logic boards, and I'm convinced that I was revisiting the PowerBook 5300c of a decade ago.

Apple now has a firmware update that reportedly solves the shutdown issues, quiets the mooing cow, and makes the machines run at sane operating temperatures. Firmware on the MacBook, like the Windows XP device drivers on the Portege and the PowerBook 5300 system enabler in System 7.5.2, caused otherwise reliable hardware to be unreliable.

So where does that leave the working professional who depends on his or her laptop? How about the student buying a back-to-school computer who needs it right now?

Lesson Learned

Simply put, let someone else buy the bleeding edge.

This is a lesson that I have learned the hard way, twice, and never again will I go against my own advice (until the next time, that is). Simply put, let someone else buy the bleeding edge. Be content with the best of last year. That way someone else can be Apple's, Toshiba's, or whoever else's beta tester, and I can enjoy reliable machines with stable, mature drivers and firmware.

My experience with the MacBook was something that I could easily have prevented by following the lessons I'd learned a decade ago.

Of course, in an ideal world companies wouldn't have to rush products like the MacBook and PowerBook 5300 to market before the software was adequately tested and stabilized. Beta testing would be done inside the company, and everything in the store would be reliable right out of the box.

We can dream. 

The MacBook Repair Saga

• MacBook pleases, but two weeks for repair is excessive, 2006.07.25. The 13" MacBook has replaced 12" and 15" PowerBooks and makes a very nice Windows gaming machine, but nobody should have to wait 14 days for Apple to fix their new computer.
• Icons, status symbols, and the MacBook, 2006.08.03. The MacBook combined the best of PowerBook and iBook designs in a fresh new way that's nearly perfect.
• MacBook repair saga: Botched and botched again, but third time's the charm, 2006.08.07. After 2 weeks at Apple, the MacBook came back running hotter than before. The first replacement MacBook ran cooler but had its own issues. The third MacBook, however, fulfills all expectations.