At Large Powell's law

At Large Powell's law

Tiring of everyone else's maxims, Matthew JC. Powell decides to invent his own . . .

To quote Winston Groom, "everyone makes mistakes. That's why they put rubber mats around spitoons." We are, each and every one of us, fallible, frail and prone to error. In case you think this is a prelude to some personal repentance or other, allow me to put your mind at ease: I'm talking about other people.

In this very publication last week, you would have seen an article claiming that Moore's Law, one of the immutable principles by which we IT folks govern our lives, may be reaching its "limit". In the same issue, you would have read an editorial by a certain editor. (I'm not saying which one. Check the panel to the right of me, we've got lots.) That editorial said, in part, that the long-cherished tradition of corporate computer buyers upgrading their equipment every three years was reaching the limit of its observance.

I'm unsure whether these notions are poppycock or balderdash, so I'll settle on pishtosh, which I think is somewhere in between. Sensitive readers, please excuse my language.

Moore's Law, stated simply, is all about microchips doubling in complexity and speed at a constant cost every two years or so. Last week's article said that this couldn't continue much longer, since current manufacturing processes were reaching the point where they could no longer produce circuits twice as small as what they could make two years ago. Let me emphasise a word for you there: current manufacturing processes. This whole industry is an ecosystem in which technological advances in one arena lead to advances in another. Moore's Law will reach its limit, and circuits will be unable to double in complexity, if manufacturing processes don't improve.

Same methods?

And what in the world makes anyone think they won't? Does Intel make chips using the same methods it used in 1968 when Gordon Moore came up with his little titbit? I think not. Once upon a time, wheels were manufactured by cavemen chipping away at chunks of rock. I've seen enough cartoons to know that. Nowadays the good people at Michelin have refined the process so that far fewer cavemen are required. So it goes with other areas of manufacturing - nothing stays the same.

Moore's law will find new limits. New processes will be discovered that will make chips so complex Gordon Moore would implode trying to understand them. The Law will continue because the limits will expand.

Likewise with the three-year upgrade treadmill. I don't know if this is anyone's Law in particular, or if it's just a tax thing, but it's been the rule as long as businesses have bought desktop computers. The opinion of my steamed editor (oh, sorry, that should have been esteemed, I guess I do make mistakes) was that he's pretty happy with the PC he's got at the moment and doesn't see that there will be any need to upgrade, even potentially years down the track. Word processors and spreadsheets don't need more power than this, he thinks, so he'll stay with what he's got into perpetuity.

I once thought like that too. As soon as I got a machine that ran at more than 200MHz, I thought "computers can't get much faster than this, and I'll certainly not be able to write any faster on a beefier machine. I'll stay with this forever". It's not quite a year since the 200MHz speedster arrived on my desk, and since then I've bought a 266MHz screamer for home use.

Guess what? I can write faster on the faster machine. It takes less time to scroll, less time to save things, less time to start an application. It's a matter of microseconds I'm talking about, here, but the machine at work just feels sluggish by comparison. That's one year later. How will I feel if I compare this machine to whatever's on offer a year from now?

One problem I have is that the software I use most often is optimised for the faster machine. The speed differential is out of proportion to the actual technical improvement in the processor. This is nothing new: software companies have always managed to drive hardware sales by requiring more than the current technology can provide. It doesn't take much tweaking of the code to make owners of older machines feel left behind, even if what they were using was state of the art just last week.

And this, gentle readers, will continue as well. Have no fear that hardware and software manufacturers will collectively decide that computers and applications run fast enough now, so there's no reason to release a new version. They will find ways to make your current top-of-the-line products passŽ.

So in conclusion, the simple statement of Powell's Law is this: whenever the IT industry approaches the limits it has set for itself, it will find a way to set new limits.

Follow Us

Join the newsletter!

Error: Please check your email address.
Show Comments