26 May, 2008

Fuel economy and Moore's Law

So gas is now $4 a gallon. I remember that, in the days following Hurricane Katrina, I found a gas station in Rocky Mount, NC that was selling gas at $3.40 a gallon. I considered myself lucky, since: (a) they had gas (many stations were empty), and (b) they were 10-20 cents cheaper than many other stations. I wonder if the Attorney General of North Carolina is still hunting down price gougers; as I recall he found all of one in 2005, and I'm not sure he even succeeded in prosecuting the guy.

There's a famous remark about the progress of computer technology known as Moore's Law: more or less, the processing power of computers should grow exponentially over time. It's generally accepted that this cannot go on for ever, but since Moore made the observation in 1965 the trend has more or less followed his prediction. So why do computers seem so slow? The major reason is probably similar to one reason that fuel economy generally declined during the 1990s.

I've grumbled before that my 1996 Saturn SL makes substantially better fuel mileage than my 2006 Saturn Ion did. Last year I traded in the Ion for a Mazda5, which has comparable mileage but is quite a bit bigger. (Having children is expensive, and not just the hospital bill.) (As for the SL, I gave it to my brother in Northern Virgina, who has saved thereby an enormous amount of money in comparison to the truck he used to drive to give singing lessons.)

You'd think that, over ten years, Saturn would have figured out how to keep the mileage at worst constant, and maybe increase it a little, but instead the mileage went down. Why so?

I've read somewhere (sorry, no link) that one reason is increased safety standards for automobiles. Some of this is government mandate, but some of it is also necessary for survival in the market. I had the impression that a lot of people looked with suspicion at Saturn's polymer (read "plastic") panels and spaceframe. Saturn was a bit defensive about it; I remember that they made a lot of noise in their car clubs and dealerships whenever someone survived a crash because of the design of the Saturn (as opposed to "in spite of the design", one imagines). I had the impression that Saturn pioneered both designs, but today you can't find polymer panels even on a new Saturn. They've become Opel USA, more or less.

What I want to get at is that these safety features increase the weight of a car. You can design better and more efficient engines all you want, but if you increase the car's weight and design it for style rather than aerodynamics, you'll lose out substantially on the mileage.

A similar situation exists with computers. People have complained that computers are slow for as long as I can remember. Today's computer are amazingly fast, to be honest, and they have a number of spectacular innovations to make them efficient: pipelining, cache, RISC influence, vector arithmetic, and that's before one talks about improvements in manufacturing and increased clock speeds.

Yet many new computers seem slower than the old ones we're used to. Largely this is due to what we're asking the computers to do. Today's computers are performing an immense amount of background computation that you don't see, and come with an enormous amount of complex software.

A good example is when Apple released Mac OSX 10.0, and when Microsoft released Vista. Lots of people complained that computers loaded with the new versions of these operating systems were dog slow.* Over the years Mac OSX has gotten much, much more usable; some of this is (again) improvements in hardware, but some of it is also due to improvements in the software. They've put a lot of effort into simplifying its complexity.

Some people have said that we could improve fuel economy drastically by switching to carbon fiber instead of steel, but none of the auto manufacturers are willing to do it because of the cost and the long time before they obtain a return on investment. Maybe one day.



*dog slow: There's an interesting phrase. I've never known a slow dog. As a matter of fact, I've never known a dog I could outrun. Even toy dogs can keep up with the bike, yipping at the tires until they tire out and trot back in satisfaction.

No comments: