Intel will start the dual-core era today with official release of dual-core Extreme Edition chips. AMD will follow with dual-core Opterons on Thursday.
Does it matter that Intel “won?”
It matters only if you think over the next three days, everyone who wants a dual-core processor will buy one from Intel, leaving AMD with no customers, and if you think that, you REALLY need to get out more.
Three days is nothing in this business, “Internet time” and those prone to hysteria notwithstanding. Believe me, this will be a trivia question three months from now, if it takes even that long.
No, dual-core will not be a multiday battle. It will be closer to a multiyear war.
But it will seriously start a lot sooner than AMD apparently thinks. Intel has a new battle plan.
To make dual-core a new standard, the following has to happen:
1) You have to learn how to make a lot of these chips fairly cheaply Realistically, this means 65nm. From current accounts, Intel has the edge here, and apparently expects to make desktop dual-core just about standard by the time AMD gets their first ones out.
2) You have to convince average people they need it A lot of geeks are getting all wound up about how little dual-core helps in most single applications. That’s completely correct, and probably will be so for the next year to two.
It’s also completely irrelevant because that’s not the only way these chips can be sold.
Intel has clearly indicated they want Joe Sixpack buying these things by next year; they expect 85% of their desktop chips to be dual-core by the end of 2006.
It is just as clear how Intel plans to get Joe to buy this: Two heads are better than one (especially when they cost almost the same). Get a smoother ride when you’re doing lots of things at once. And did we say it costs little (or maybe by next year, no) more than a single-CPU?
People think AMD is a price innovator, but they forget that Intel has a long tradition of pushing new technologies ahead of old by simply charging the same price for both.
Joe will look at this and say to himself or herself, “Two has to be better than one, so I’m going to get two for the price of one.” That will be the core of the Intel marketing strategy. And do you know what? It’s probably going to work, because for most Joes, they’ll be right.
Spare me the benchmarks, I know better than most that AMD will win those, but that doesn’t matter to the average Joe. A dual-core is likely to give the average Joe a better computing experience for what he does than a fast single-core. Joe won’t notice Word or IE open up a split-split second faster. Joe does notice when the machine stalls as it shifts from one activity to another. Promise him a machine that doesn’t do that, for little to no more money, and he’s going to like that (or at least he will after enough ding-dong ads).
This is what virtually all the computer hardware people (and, not so incidentally, AMD) don’t get. Dual-cores are not aimed at geeks; they’re aimed at Joes.
Some may say, “This is all Intel can do with its lemons.”
Intel is simply taking their lemons and making lemonade from them. This is a very smart thing for them to do.
So while AMD tries to get megabucks for their server dual cores, stalls on dual core desktops to keep them from cannibalizing Opteron sales (or even general capacity), and tells people single-cores are still the way to go; Intel will stoop to conquer the vast majority of the market under the geek radar.
Don’t think that will happen? Look at Prescott. To put it mildly, it hasn’t received too many rave reviews from the computer hardware press. Have Intel sales gone to hell? Hell, no.
That’s one big difference between AMD and Intel. AMD thinks it has to impress geeks; Intel knows it has to impress Joes, and Intel is big and powerful enough to not only play the game, but change the rules.
True, sometimes Intel tries to change the rules and fails badly. However, when they’ve failed, they’ve failed because they’ve tried to foist something much more expensive onto the computing population. They aren’t making that mistake with the “regular” desktop dual cores; they won’t cost much more than singles.
Nor do dual-cores have to be an overnight success on the desktop. Intel wouldn’t want that until 65nm. The 90nm duallies are merely meant to establish a beachhead for the full assault next year (and give Intel something to talk about this year).
A Lower Priority
There will be one last phase in the dual core wars:
3) Increase the use of multithreading in single applications The problem geekdom (and AMD) has is that they think this is the first (and probably only) priority. It isn’t. While it isn’t the last priority, it will be the last to show up, and frankly will be a mopping up action, not the decisive battle.
If history is an indicator, one of two things will happen with multithreading. It will either start to kick in seriously for apps that can really use it about eighteen months from now (like with SSE/SSE2), or it won’t seriously kick in at all (like MMX).
The first is more likely to happen than the second, though a majority of programs probably will never seriously go multithread simply because it doesn’t help that program.
What is important to realize is that multithreading might well be important to you in determining what kind of computer to buy, but you (and the rest of geekdom) aren’t the world.
The problem for AMD is that they give every indication that they’re going to fight the last war, while Intel will be busy setting up and fighting the new one. They’ll make enough 90nm dual cores and take the price/capacity hit to get the market ready for dual-core, then make them the mainstream standard once they have 65nm down. leaving AMD behind and whining for most of 2006.
AMD will emphasize performance, gain a few market share points for a while, then give it back once the 65nm dual core bandwagon kicks in and people start asking AMD sellers, “Why can’t I get two from you at a reasonable price?” By the time they have their 65nm dual core, Intel will have already established a dominant position here.
In other words, Intel’s strategy is “You take the geeks, we’ll take the rest.”
Two People, Two Solutions
The point of this article is not to tell you to buy an Intel dual-core rather than a Hammer. For most reading this, the Hammer will be the better bet.
The point of the article is to suggest that what might be best for you will not be best for those with different habits and goals.
To give an example, three or six or twelve months from now, if you’re a hot-shot gamer, a Hammer box will probably be the way to go, but a dual-core is likely to be the best bet for your CD-burning, DVD-watching non-geeky friend.
The war continues, but it keeps morphing into different forms. Yesterday’s causes won’t be tomorrow’s.