On The Sidelines . . .

Add Your Comments

If somebody doesn’t at least start to get ready for next decade’s big CPU game, somebody isn’t going to be in any shape to play.


VGA Trader If you’re interested in trading or selling your video card one of these days, you may want to bookmark this website.

Intel recently provided a presentation to indicate where it plans to go with Atom-class processors.

It’s obvious Intel finds tinyPCs to be the next big thing in computing, and just as obvious that this is a long-term project. Atom is merely warmup, with its first “real” tinyCPU chip, a system on a chip, not showing up until 2010, with true powerhouses in a phone not showing up for a few years after that.

Intel thinks it can have up to a billion transistors in one of these chips by 2013, which will certainly be more than enough to do just about anything you can do today.

It’s a major, major effort, and Intel is well aware that it might end up destroying most of its current business in the process. The payoff is creating a market that eventually will demand billions rather than hundreds of millions of chip, stuffed into anything that could use computing, and all using variations on the same simple, cheap building blocks.

That is Intel’s vision, One Ring To Rule Them All. It is truly visionary, though you might call it nightmarish.

If you think it’s the latter, what you have to ask is, “Who is going to stop them?”

Normally, the first and sole answer would be AMD, and they’ve said both yes and no to this challenge. About a month ago, they said that they would come up with an Atom competitor in the form of a stripped-down Athlon 64 core.

At that time, we said that cannibalizing what you already had was simply not going to cut it as a strategy in this new world.

Now AMD is backing off from even that, saying that it would sit and wait and see how things develop, saying in essence: we can’t afford to do everything.

It makes little difference whether or not AMD comes up with an Atom competitor. Atom is just a stretching exercise before the game for Intel.

What is critical for AMD’s long-term viability or even survival is that they must have a player fully equipped to play this new game when the real game starts around 2010.

What is critical for the world to understand is that it’s not just a matter of cutting down a current CPU or even drawing up a new CPU design. Creating a tinyCPU powerful enough to approach today’s performance standards, but able to work within the power and other confines of a smartphone isn’t just a matter of process shrinks. It’s not even quite a matter of building a chip from the ground up. You have to literally think in ways that are not like, or even the opposite of typical CPU design, and you need a corporate design architecture culture attuned to that.

It’s like taking a bunch of people from their well-watered worlds, dumping them in the middle of the Sahara and telling them, “Fend for yourselves.” Yes, Bedouins can live in the desert, but that’s because they’ve practiced a way of life since birth focused on making every drop of water count.

Think you could all turn into Bedouins just like that?

Since the development of the initial Pentium-M, and to a much greater extent since the Prescott fiasco, Intel has been growing a mainstream culture of designers and engineers obsessed with saving power, one tiny little bit at a time. Besides maybe a few outside the AMD mainstream working on Geodes, you don’t have that kind of power-miser culture at AMD, and you’re not going to create one instantly when you need it.

I don’t doubt current AMD engineers leavened by a few outside hires couldn’t come up with 80% of the power savings a group of true power misers would get, just as that stranded group in the Sahara could cut down on most of their current water use with the help of a book or two on the subject.

But in the world of tinyCPUs, that 20% difference is going to be the winning edge. That proved to be true for notebooks, because people wanted more battery life for their weight, and would be even truer with tinyCPUs.

Deciding Between When You Should Do Both

This sounds bad, but it’s actually overoptimistic. AMD would start well behind Intel if they launched a true tinyCPU project today, but at least they’d be close to having something by the time of the first regular season game.

The real problem is AMD isn’t starting a serious tinyCPU project today, or tomorrow, or next month, or maybe even next year.

They can’t afford it. They can’t afford the money or engineer time right now; more immediate needs and payoffs take higher priority. This is a huge long-term strategic error, but those saying “No” to funding tinyCPU developers are probably saying, probably quite rightly, “If we don’t do the other things first, there isn’t going to be an AMD to make tiny- or any other kind of CPUs.”

They can’t afford to do it, and they can’t afford not to do it.

This is the tragedy of lack of money at AMD. This is why finances matter even to those who just want to buy CPUs so they can play games. The CPU battles in the CPU wars aren’t won when the products come out, they’re won and lost behind the scenes in the research and development stages years before the products come out.

Look at Intel. They have money, enough money that they can gamble some and afford to lose it. This Atom project is just starting. If there’s going to be a big payoff, it won’t come for another four or five years. But Intel can afford to be patient, and it’s not like this project will bankrupt them if things don’t work out as planned. Look at Itanium.

In contrast, AMD can’t do that. On the whole, it’s not that the AMD decision-makers are much dumber than those at Intel. Money is the key difference. Even in better days, AMD has often been forced due to lack of money to do A before B when I’m sure they knew they really should have done both.

And these days, the decisions are more like “Die now or die later” and tinyCPUs are definitely in the “die later” category.

However, when you put off doing something until it cannot be put off any longer, you end up with crisis-driven crash projects. You start gambling and cutting corners to try to meet insane deadlines, and that means you’re very likely to screw up. Barcelona is just the most recent example of an AMD crash project that crashed.

It’s not that Intel is immune to this, the ultimate failure of the PIII and PIV lines happened because they decided, “Fixing our leaky transistors will hurt our profits too much, so beef up the power and the fans.” But these disasters are more likely to happen when you don’t have money.

Even if AMD fails in the tinyCPU arena, or comes up with too little, too late, Intel will not win by default, the stakes are too high. At the least, some sort of consortium centered around ARM, nVidia and no doubt others will give Intel a fight. Rest assured, the fight over tinyCPUs will be THE CPU fight next decade.

But AMD can’t seriously fight that war if they don’t start seriously preparing for it soon.

Ed

Leave a Reply

Your email address will not be published. Required fields are marked *