“The stone which the builders rejected has become the cornerstone.”–Mark 12:10
In April, AMD wouldn’t even talk about it; they were too busy Hammering away. Just in the past handful of months, whether it would even see the light of day was put into doubt.
In six months, Barton goes from dead dog to underdog to topdog on the AMD desktop, now that AMD has clawed back Hammer deployment.
Can The Understudy Cut It?
The primary problem with Barton is not so much it taking the limelight, but when.
Putting the understudy in after the star broke a leg is one thing, but Barton showing up around March is like the understudy showing up at 9 for an 8:00 show. It’s just not a crowd-pleasing move.
Had this come out in time for Christmas sales, Barton would have been a creditable answer to the 3.06GHz PIV, and could have more-or-less held the fort until Intel came out with Prescott.
Coming in March, well, that dramatically cuts back on the time in the limelight. Prescott may show up as early as June (though I think September is more likely).
Perhaps more importantly from the financial end, Intel should at the least be able to get a 3.2GHz chip out by around then, which means they’ll be able to knock prices down on 2.8GHz and below and keep AMD in the price trap.
Sure, it will be a great final upgrade for those married to socket A, but for whom else?
Can Anyone Here Play This Game?
What’s more disturbing about all this is that AMD seems to be running around like a chicken with its head cut off. Building CPUs is a not spur-of-the-moment kind of activity. It takes preparation and time to shift production.
Let’s take these Bartons. Initially, they were only supposed to be made at UMC (and starting newbies off with that never made much sense). Is that the case now? Who knows?
Will Dresden make them? I can imagine some poor fellow working at Dresden hearing over the last six months, “You’re going to be making Hammers, no, TBreds, no, Bartons (and perhaps also soon if not already, “You aren’t going to be making anything at all; we’re letting you go.”)
How much time, effort and money is getting wasted by these constant shifts? Don’t you wonder at least a little if life at AMD is like, “Let’s make TBreds! Oops, we can’t make them right. Let’s try Hammers instead. Oops, can’t make them right, either. How about SOI? Oops, 0-for-3. Next!”?
Imagine you’re a Taiwanese mobo manufacturer. For most of a year, you at least had been designing and perhaps beginning to gear up manufacturing for Clawhammer. Now you’re told literally, “Wait until next year.” You cannot be happy about this.
Does Clawhammer Even Make Sense As A .13 Micron Chip Anymore?
.09 micron Prescotts will definitely be out by the time Clawhammers get made.
My guess is you’re not going to get more than 2.5GHz out of .13 Hammers. That could justify a PR of 4000+ (indeed AMD has never suggested anything more than that for .13 micron), but given Prescott, that gives you a one-hit wonder.
By then, 256K cache will look awfully skimpy for an allegedly leading-edge processor. 512K is already standard on Northwoods, and Prescotts will come with 1Mb. Increase cache, and Clawhammer isn’t so small anymore.
Even worse, single-channel DDR I memory will be antiquated by fall 2003. Dual-channel will be mainstream by then, with DDR-II’s arrival imminent. What do you do with that integrated memory controller? Design it for dual-channel? DRR-II? Both?
All these combined means going back to the drawing board big time, and it might make more sense to try to do all this in .09 micron six months later rather than delivering a bloated, stressed-out CPU at .13 micron.
Or will AMD make the same mistake with Clawhammer at .13 than Intel made with Willamette at .18?
Given their recent track record, you have to at least suspect that AMD will tell you everything is wonderful with Clawhammer until about October 2003, then realize all the items I’m mentioning in October 2002, and oops, there it goes again.
Fool Me Once. . . .
I mean, really, how much confidence do you have left that when AMD says they’re going to do something, that they’ll actually do what they said they were going to do when they said they were going to do it? It has to be at least shaken by now.
And neither you nor I have any real stake in the company. Imagine having a lot of your money or your company’s based on that word.
Or how about sticking your neck out and saying your company should buy a bunch of Opteron servers against the conventional wisdom of the Intel zealots on your server staff? If your boss says to you, “If AMD becomes history, so will you.” what do you say?
Even in this day and age, assets are not just measured in buildings and equipment and cash in the bank. The value of the faith customers have in the reliability of a company can’t be measured in dollars and cents, at least not until it’s too late.
There’s even an accounting term for it. It’s called goodwill. It’s that which makes a company worth more than the sum of its parts. When customers view a company not like a rock but more like sand dunes, both they and it just erode away.
And AMD has been depleting their reservior of customer goodwill faster than their cash balances.