There’s some dispute about whether Clawhammers will or will not show up Q1 2003.
The AMD CC says one thing, which we reported yesterday.
Again, please note that AMD PR never says that Clawhammer will be out at end of Q1, but rather “Hammer” or “64-bit” systems. Again, Opteron workstations have long been in AMD’s plans, and such a thing showing up next March or April would be consistent with both the CC and PR statements.
Finally, there’s a roadmap over at Tom’s Hardware which indicates a >3000+ Clawhammer for Q1 2003.
However, this is an outdated roadmap. For one thing, it does not indicate any 333MHz FSB processors. Nor is the TBred 2700+ there (which AMD indicated October 1 would be a mass production product; only the 2800+ is supposed to be a “limited edition product.”)
If AMD PR would actually like to straighten this out, all they have to do is say exactly when Clawhammer is supposed to go into initial production and in what kind of volume.
For right now, it looks like AMD plans on having Sledgehammer servers and workstation available next spring, and that there will be few if any Clawhammers around.
We note in passing that neither the AMD CC nor AMD PR mentions Clawhammer or Athlon Hammer by name. Whether this omission means anything remains to be seen.
Some More Spin
In the recent past, I’ve heard two other explanations for AMD’s behavior that at first glance may seem reasonable, but isn’t after a closer look.
The first is “We have doubts the market wants to buy high speed processors.” That was said a couple times during the AMD CC.
This is of course complete nonsense if you think about it a moment. What is true is that the market is less and less willing to buy high-priced processors, which is something quite different.
Whoever said, “Give me a slower processor for the same amount of money?”
The trend towards lower-priced processors has certainly not stopped Intel from coming out with 2.4GHz and better processors, in fact, they accelerated ramping a bit over the course of this year.
You can buy all the 2.4 and 2.53GHz PIVs you want from Intel at fairly reasonable prices; so keeping high-speed Athlons scarce and expensive makes no sense. Besides, AMD makes more money selling two 2400+s at $130 than one at $200.
What may be much closer to the truth is that the six-month long Palomino liquidation sale will make it even tougher for AMD to get a good price on any of its chips.
Besides, AMD is bound to run out of Palominos someday. What else are they going to sell? AMD’s own datasheets show TBred2s starting at the 2000+ level. If Palominos really are almost all gone, they just start selling TBreds2 starting at 2000+ (or maybe a bit less) instead. That’s just normal progression. Unless, of course, they can’t really make them.
AMD’s actions belie their concerns about selling high-speed processors. If things are so rough out there, why then do they plan on having as many as four different processor lines running at more than 2500+ out there fairly shortly?
Really. Count them:
TBred: 2600+, 2700+
Barton: around 2700+-3000+
Opteron Hammer: 3000+
Athlon Hammer (if available): 3000+
This could well be overly nitpicky, but the AMD CC said that AMD “can make” two million 2400+s this quarter. This is a rather odd statement when you remember that Dresden was making at least double that just six months ago. If you’re not making Palominos anymore, and you’re not making “Hammers” or Bartons yet, then what else is there for you to make?
The second line I’ve seen is “It’s not the right time for Hammer.”
Sorry, but there is no good external reason for AMD delaying a finished product in the CPU market. These are CPUs, not wines. They don’t become more valuable the longer you hold on to them, they become less so. Nor can you make any money not selling products you don’t make.
Even if that weren’t the case, AMD doesn’t have any real ability to independently determine pricing, anyway; Intel does. They effectively set the maximum price for a CPU with their pricing, and AMD sells at a discount to that. All AMD has been able to do in the past to affect pricing is increase the discount and hope to drag Intel along with them.
In theory, AMD would be freer to set its own pricing on items that are better than Intel’s best (though that didn’t prove to be the case when AMD did have a lead with the Thunderbirds), but again, something like Clawhammer isn’t going to improve with age simply because Intel will continue to progress. A 3400+ Clawhammer may lead the pack in March; it won’t in September.
There can, however, be good internal reasons to delay. The problem is all those “good” reasons are due to bad causes. If you just can’t crank these things out easily and effectively, or if you don’t have the money to do the necessary plant conversions right away, then, yes, it does make sense to put what you can make into higher-priced markets like servers.
The problem gets worse if Clawhammer can only be made in bulk as a 3000+ processor. With that kind of rating, it can’t claim to be faster than Intel’s best. It can’t even claim to be appreciably faster than AMD’s second best.
Unlike Sledgehammer, which has additional features and benefits and is being sold to a virgin market. Clawhammer has to compete against very competitively priced and featured Intel and AMD processors. Why buy an expensive Clawhammer AND mobo for negligible improvement over just popping in a Barton?
It would be hard enough to wean people away from socket A with a significant superior product. If you can’t mass produce a Clawhammer significantly superior to either the PIV or Athlon/Barton, best not to even try until you can. That is far more likely to be the real reason for delaying Clawhammer.
Pure speculation, but don’t be shocked if Clawhammer doesn’t come out in significant numbers in .13 micron, or even comes out at all; AMD might do an Intel, stick with the Sledgehammer design and perhaps castrate some cache (or not) and call it a Clawhammer. There’s some good reasons not to do that, but a .13 micron Sledgehammer on the desktop would offer more competition against Prescott than the current Clawhammer design.
But doing the best you can do under poor circumstances is not the same as doing well.
What I Think Is Going On
AMD certainly did have major problems with TBred2, and probably continues to have at least some production constraints, but they can probably get a significant but not plentiful numbers of TBred2s out this quarter.
These constraints may be due to yield, or perhaps more likely due to keeping a good chunk of Dresden capacity down to regear for Barton/Hammer.
AMD’s actions with Hammer seem to be symptomatic of yield problems. The most likely major cause of that is SOI. Under those circumstances, selling low-yield Opterons into what hopefully will be a high-priced low volume server/workstation market makes sense. Provided they can sell what they can make in those markets (which is questionable), holding off on mass production of less profitable Clawhammers until yields go up makes sense.
AMD will probably end up stretching out the conversion to Hammer process at Dresden to cut capital expenditures, and again, if you have to delay, Clawhammer is what you delay.
I Just Want Facts
What I’m doing here is like getting a crossword puzzle with most of the pieces missing. Essentially, I try to put myself in AMD’s shoes. Would I take these actions if the situation as described by them were true? No? Then what circumstances would make me take such actions?
It’s not that I think the actions being taken make no sense, it’s that the explanations for those actions make no sense. You just wouldn’t take these actions if everything were going well. You’d only do them if they weren’t, and these were the best choices you could make under poor circumstances.
If Intel said tomorrow, “We’re postponing Prescott for a year because we’re doing so well and don’t need it,” would you believe them, or would you think, “Intel has some kind of problem here?” That’s what we have here.
So I try to come up with explanations for the actions based on the facts available that do make sense.
Yes, it would be better if I had all the pieces. True, I could be missing vital pieces of the puzzle and get the wrong picture.
But what’s better, trying to figure it out, or just taking somebody’s word that the puzzle is really the Mona Lisa even though none of the pieces match the right colors for that painting?
All the pieces of the puzzle just aren’t out there. That’s because AMD is the North Korea of CPU companies. If you want to believe anything and everything they say, I have a bridge I can sell you.
The problem with “I just want facts” is that this can be turned against you. If you don’t want to look at independent, skeptical analysis, all a company has to do is cut real information off, and then you’re at the mercy of the PR team. If you’re so concerned about bias, who could possibly be more biased than people paid to be as biased as possible for a company?
Doing that isn’t smart, that’s being a fanboy, whether intentionally or not. It’s no different from Macfanatics believing Apple when they say the Mac is much much faster than a Pentium in some weird, totally unrepresentative benchmark
I wish it weren’t necessary for me to try to read tea leaves, but blame the place that’s withholding the hard facts, then spinning the couple they do put out beyond all reason, not those trying to make some sense out of it.
What’s Your Real Reason For Saying This?
There’s an old principle called “Occam’s Razor.” It basically says, “The simplest explanation is usually the correct one.”
I never cease to be amazed after writing an article giving forty-two reasons why I think something, then someone writes me and asks, “But what’s the real reason?” Uhhh, try looking at the article.
It’s like a doctor telling someone, “Your mother died last night.” The person replies, “Why do you think that?” The doctor says, “She isn’t breathing, her heart isn’t beating, she has no pulse, and her body is cold and stiff.” Then the person says, “I think you just don’t like me. What’s your real reason for saying that?”
Why am I writing articles saying AMD is in trouble and doubting the company line? Because I think AMD is in trouble, I don’t believe the company line for all the reasons I state, and I think that’s an important story to cover.
Just like a couple years ago, when Intel ran through a time when they couldn’t do anything right, I said Intel was screwing up, too, and if they kept that up for a few more years, they’d be in trouble, too.
I’m not against AMD. I’m not against Intel. I’m against screwing up and not being honest to your customers and investors about it.
Fault the accuracy of the facts or fault the reasoning, but don’t think everything and everyone in the world is like The Matrix. Judge the argument, not the arguer.