Avoidable Contact #57: Tragic failures, and tremendous triumphs, of the tyrannous clean sheet
“According to scientists, the bumblebee shouldn’t be able to fly.” Surely you’ve heard this. It’s a seductive phrase, because it allows us to express our misgivings regarding Science-With-A-Capital-S. Those misgivings are often well-founded—the “reproducibility problem” in modern science is a genuine elephant in the room, and many studies are published with great fanfare only to be silently withdrawn shortly afterwards—but the flight of the bumblebee shouldn’t be used as a metaphor for them. Bumblebees fly just fine. They do it by flapping their wings forward and back instead of up and down. A few scientists correctly noted decades ago that there wasn’t enough wing surface on a bee to support traditional wing-powered flight, but that’s like saying that there isn’t enough flipper surface on a human body to let us swim.
The modern automobile is no less an apparent miracle than a bumblebee. It shouldn’t be able to perform flawlessly for years in all sorts of weather, returning astounding gas mileage and providing astonishing audio quality, while receiving little to no maintenance. Eppur si muove, however—and yet it moves. This didn’t astound me as a teenager, because I had not yet entered the delightful and occasionally profitable world of software development. It astounds me today, because I’ve spent two decades watching people who are supposed to be the cream of the human crop fail to accomplish the simplest tasks over and over. If the best and brightest among us regularly fail at mundane tasks like lining up every field in a data-entry form, how is it that you can spend $19,999 at a local car dealer and get an actual physical product which will drive you six times around the world without a valve adjustment and save your life in any crash you encounter during said drive?
Turns out there’s an easy analogy for those of us who grew up swinging a compiler instead of a Snap-On socket: the so-called “x86 architecture,” pioneered by Intel and widely copied elsewhere. I encountered it for the first time in 1986, programming an 80286-equipped IBM PC XT in WATCOM FORTRAN at Theeeeeeeee (that’s how they pronounce it) Ohio State University between my sophomore and junior years of high school. I’m using it right now on a Core i7-powered Lenovo Ideapad Y900, an old battleaxe of a gaming laptop which has real mechanical switches in its keyboard. The processors in those two computers are in no way interchangeable; it’s like comparing a Benz Patent-Motorwagen to an S63 AMG. There is, however, a clear line of heredity and succession which connects my 1986 computer to my bought-in-2018 one, the same way there’s a line of heredity and succession between the first Benz and the latest one.
It’s worth noting that the x86 processors have pretty much never been the hottest thing on the block. The early versions were often beaten out of their socks by variants of the consumer-focused 6502; look at a Super Nintendo game against a PC game of the time for an extreme example. Motorola made faster and better chips for a long time in the consumer-computer market; on the business side, the ULTRASparc, PowerPC, and HP’s PA-RISC processors offered more elegant programming solutions and higher raw performance. Didn’t matter in the long run. The X86 thrived because it was a consistent and reliable performer with a lot of support from manufacturers. A few pretty high-profile problems (FDIV bug, anyone?) didn’t derail that train. Intel (and their same-but-different competitor, AMD) offered a steady diet of improvements in the x86 world. Any major changes (and there were some) tended to be hidden beneath the hood, so to speak. Today’s top computer processor, the evocatively-named Ryzen Threadripper, can trace its roots directly to 1978.
Like computer processors, the best and most successful cars are usually enhancements of, or incremental improvements on, their predecessors. Here’s an example: In 1977, General Motors “bet the company” on new editions of their full-sized B-body and C-body sedans. Had the new cars been failures, the fallout could have led to bankruptcy or worse. History tells us that the new-for-1977 B-bodies were a massive success—gorgeous, spacious, reliable, capable.
Three years later, GM introduced the front-wheel-drive X-body, which was arguably just as critical as the B-body rollout three years prior. This time, however, the Chevrolet Citation and its siblings became infamous for deadly brake failures and catastrophic quality control. Hundreds of thousands of customers abandoned GM as a consequence, putting the company into a tailspin that would eventually see it all but abandon the passenger-automobile market in favor of trucks and SUVs.
How did the X differ from the B? It’s simple: The X-car was a clean-sheet design, while the B-body was actually a thorough restyle and repackage of the “Colonnade” A-body intermediate car from 1973. Having the well-respected and thoroughly-debugged A-body frame beneath the 1977 full-sizers meant GM could concentrate on what mattered. The company also had access to some of the most time-proven engines in the business, from the fabulous Cadillac 425 to a pair of 3.8-liter V-6es, one of which was simply a cut-down variant of the small-block Chevy.
The X-car, on the other hand, was a completely new thing; everything but the “Iron Duke” four-cylinder was a clean-sheet design, and even the Duke had to be revamped for transverse duty. This clean-sheet approach allowed for some astounding engineering feats—the original 5-door hatches are marvels of packaging efficiency even by modern standards—but it also meant that pretty much every aspect of the car would be undergoing a baptism of real-world fire at the same time. Some of those aspects turned out to be unreliable. Others turned out to be dangerous. The A-cars (Celebrity, Ciera) which arrived two years later were much better vehicles, largely because they had the benefit of hard-won engineering data. It was too late for many customers, who decided to cast their lots with Toyota or Honda instead.
Don’t be so fast to laugh, GM haters and import lovers; the X-body is far from the only example of catastrophic clean-sheet engineering in the automotive world. Toyota’s early efforts at front-wheel-drive were so dire that the company kept a driveshaft in its core-product, must-not-fail Corolla long past the time when the rest of the competition had gone transverse. When Jaguar completely reimagined the XJ6 for 1988, creating their first clean-sheet car since 1969, the result was a beautiful, powerful, wickedly competent sedan… which all but required a ride-along mechanic. Mercedes-Benz made massive changes in all three of their sedan lines near the end of the Nineties, creating three model numbers (W203, W210, W220) which even today strike terror in the hearts of previous owners. Let’s not forget Porsche; terrified by European Union drive-by noise regulations, the firm threw away a basically perfect car (the 993) in favor of the 986/996 platform. The first few years of those cars had more issues than National Geographic, to put it mildly.
Nearly every clean-sheet design of the past four decades has been, shall we say, troubled in its first few years. Heck, not even Rolls-Royce could get it right. The first Silver Shadows were beset with reliability problems. Not until the Bentley Turbo R, which was basically a 26-year-old facelift of the Shadow, did Crewe really get it right. That’s a long time.
There are exceptions which prove the rule. The C5 Corvette was a clean-sheet design developed on a shoestring which proved to be fast, reliable, and desirable from Day One—but it used many ideas from the C4, which had been even more of a fresh idea and which of course had a few bad years up front. Porsche’s 991 was a decent car right away, but it used the engine from the 997.2 and many of the (previously troublesome) electronics from the company’s trucks and sedans. I’m sure some of my readers can think of other examples.
Sometimes a clean-sheet vehicle appears to satisfy all expectations, only to display a fatal flaw down the road. Ford’s new-for-1997 F-150 was a winner right away, but the crash performance turned out to be pretty dire. Honda’s move to a fresh global platform for sedans, SUVs, and minivans seemed to go off without a hitch—unless you happened to like the combination of V-6 and automatic transmission.
It’s a common hobby of automotive journalists to express a sort of childlike wonder at the continuing success of the Dodge Challenger. OMG, BECKY, IT’S SO OLD. Certain parts of it are even older than you’d suspect; I’ve been told by Mopar experts that the front suspension shares certain critical dimensions with the 1986 Renault 21 (known here as the Eagle Medallion), in much the same way that the 1996 Dodge Stratus was derived from the Neon which was derived from the Sundance which in turn was derived from the 1978 Dodge Omni. Certainly there’s nothing particularly new-feeling about the Challenger. Yet it’s more popular now than it’s ever been. What’s the appeal?
The Challenger looks like a million bucks, particularly in Hellcat Redeye trim, but I also think that today’s consumers have stopped being interested in newness for the sake of newness. We are keeping our cars an awfully long time nowadays. If you’ll still be paying on your car 71 months from now, what do you care if it was designed in 2018, 2008, or even 1998? Why not get a known quantity? Thus, the success of the thoroughly-proven Challenger. No clean sheets in this thing. Like the bumblebee, it’s big and heavy and ungainly-looking and just a little threatening-looking. Automotive scientists say it can’t fly out of showrooms—but it does.