Does Anybody Want This CPU?

XBitLabs reports some shocking Hammer production projections.

Condensed, here’s what it says:

CPU Type
2Q 2004
3Q 2004
4Q 2004
1Q 2005
Socket 754
Socket 939

Before we go for the jugular, we’ll note that AMD seems to want to pump socket 754 for a few quarters, then suddenly turn around and dump it. This ought to make those interested now in socket 754 systems think twice about it.

But that’s not what is most important here.

What is important is that AMD doesn’t plan to make many Hammers over the next year, period. Even a year from now, desktop Hammer production will be less than 20% of a decent production quarter for Dresden today.

Remember, a year from now, Dresden should be pretty much completely converted to 90nm, Hammer die sizes will be around the same as XP/Bartons are today.

This is not a normal ramp. For the sixth full quarter after introduction, the proportion of new generation chips being produced is rather less than you would normally expect by the second quarter after intro. Even if you factor in all the reasons for delay, the figure for Q1 2005 is half or less what you might reasonable expect.

Let’s take the projected production numbers being cited today for Q2 2004 and Q1 2005 with the estimates that were made for Q2 2004 just six months ago:

CPU Type
New 2Q 2004
New 1Q 2005
Old 2Q 2004
Socket 754
Socket 939

Big difference, isn’t there?

There’s something wrong somewhere here. You can come up with at least decent excuses for Q2 and Q3, but those excuses get shaky in Q4, and really fall on their face in Q1.

Please also keep in mind that the projected production figures are little higher than AMD’s stated current production capacity for Hammers. AMD said it can make substantially more than a million Hammers a quarter today.

Either AMD can’t make them, won’t make them, can’t sell them, or some combo of these three.

It’s Not Like They Have Anything Better To Do

A year from now, what else is AMD going to be selling? What is going to pay the bills? These Hammer figures aren’t enough to sustain the company, and even if they were, it would mean a dramatic drop in market share: these numbers represent only 3-4% of world wide CPU sales.

I really don’t think that’s the case, but what else are they going to do?

Athlon XPs and Barton? How well could they be selling a year from now? At what price? And finally, why would you rather sell them than Hammers?

What will they even look like a year from now? Allegedly, AMD is converting Dresden over to 90nm. To what degree? Is AMD just converting a little to make the Hammers due to financial constraints while leaving most of Dresden making 130nm XPs?

Or has AMD changed its mind, decided they’re going to make 90nm XPs, and just won’t tell anybody yet? I’m not saying that’s the case, but when you see the numbers for socket 754 processors plummet in Q1 2005 just when the budget socket 754 replacement should be really kicking in, it makes you wonder.

Are they making Opterons instead? Please. AMD would have to capture just about the entire server market to fill the production gap. If you believe that’s going to happen, you probably still believe in Santa Claus and the Tooth Fairy, too.

Opterons will do well over the next year, but going from tens of thousands to millions wouldn’t be successful, it would be miraculous. One should never bet companies on miracles. 🙂

The Endless Enigma

Some may say these numbers are worthless since they’re not certified coming from AMD (they’re probably coming from some mobo manufacturer). This may be, but the real problem is that certified AMD numbers would be just about as worthless.

This is a big problem for AMD. It may not seem like a big problem to someone buying a system, but if you’re running a company that is dependent on what AMD does (say, a mobo company); it must be a nightmare. You don’t know when to make equipment, and more importantly, how many.

Just to use the numbers mentioned in this article, six months ago, AMD essentially said, “Get ready, we’re ramping up big time.” So you get ready for that, and a few months later, it’s “Never mind.”

When a company does that to you enough times, you start taking their estimates with a grain or bushel of salt and minimize your commitments. Of course, when you do that, when the dam does break, you aren’t ready, and that causes further delays.

It’s probably safe to say that OEM lack of interest is at least a good-sized chunk of the problem, but what’s the problem? Are they just not interested at all? Is it a pricing problem? Is it “we just want cheap XPs?” Is it “we’re waiting for 64-bit OSs”, or is it a production capacity problem? Who knows?

Whatever it is, it’s just not normal for a new generation CPU to have such a small proportion of sales so long after introduction.

All this may seem academic to you, but what happens behind the scenes determines whether or not AMD can capitalize on Intel’s increasingly evident problems.

And numbers like these indicate that they don’t expect to.


Be the first to comment

Leave a Reply