It’s Not A Roadmap; It’s the Daily Racing Form
You can see the roadmap presented by AMD at the Morgan Stanley Dean Witter Semiconductor and Systems Conference here.
The roadmap is slide 20.
It says the following about Palomino:
Production 1Q ’01
It shows Palomino as both a “Desktop” and a “1-2P Server and Workstation.” There is absolutely no doubt that roadmap calls Palomino a desktop CPU.
Now look at the roadmap AMD put on its website.
We get told a much different story.
First, there are no more desktop/server/workstation Palominos with samples 12/00 and production 1Q ’01. The only remaining horsey with that designation are the mobile Palamino.
Second, where there was once one Palmino in the non-mobile crowd, now there are two.
The first is a 1-2P Server and Workstation Palomino, with no speed designation, sampling 1Q ’01 and going into production 2Q ’01.
The other is a Desktop Palomino at better than 1.5Ghz sampling 2Q ’01 and going into production 3Q ’01.
Since AMD hadn’t seen any difference between a “server and workstation” Palomino and a “desktop” Palomino two weeks ago, I doubt there’s any difference between the two now.
In any case, neither of these are in production yet, according to the roadmap. Only the mobiles. Seems pretty clear any horses running around right now are supposed to go into notebooks. Problem is the notebook people aren’t ready for them. See here and here.
Not Enough Horses To Go Around?
So AMD is stampeding its horses to a market that can’t use them yet.
The only way this makes sense is if the Dresden stud farm isn’t putting out enough yet.
There’s two possible reasons for this:
1) The horses aren’t coming out right and/or
2) They’re too busy making birds, TBirds, that is.
As to the first, if you look at the CNet article with a jaundiced eye, you see less than a flatout denial of any problems.
For instance, to quote from the CNet article:
“The fundamental technology is sound,” one computer executive said of Palomino, speaking on the condition of anonymity.”
Saying that the fundamental technology is sound implies that the nonfundamental technology isn’t, at least not yet. Otherwise, why even use the term “fundamental?” Just say the technology is sound.
“Reducing the energy consumption has been one of the major engineering tasks. To that end, AMD has tweaked circuitry on both chips. So far, neither sources at AMD nor within the PC industry have indicated that AMD has not succeeded in this project.”
Not exactly verification. Saying nothing means just that: nothing.
AMD has one place to make these things, Dresden. Right now, they’re making as many TBirds as they can. Shifting to a different core is bound to disrupt production a bit.
I suspect, though, not as much as it would, say, Intel. If AMD is known for anything, it’s known for changing manufacturing processes at the drop of a hat, and I would bet as new “tweaks” prove themselves, they get incorporated into new designs.
Nonetheless, there’s no reason to target chips to particular markets if you can make enough for all markets. So there is at least a manufacturing problem at Dresden. May be just a matter of converting production over, may be more than that, but it’s something.
You Bring Your Best Horse To The Horse Show
My suspicion is that AMD at least had a lot more problems with this than they’ve let on. Their actions indicate that they either just fixed them, or they are still having problems; after all, the mobile Palominos are meant to gallop at much lower speeds than any “server” or “desktop” pony.
After the notebooks are supposed to come the “server/workstations.” There’s an article mentioning dual-Palomino workstations run by AMD. Interesting discrepancy between it and the single-CPU Palomino demonstration. The single Palomino is running at 1533Mhz, while the dual is running at just 1200Mhz. Add to that another dual-display demo being run by VA Linux (presumably with CPUs a bit older than those used by AMD); that’s only being run at 900Mhz. Hmmmmm.
Let Us Pretend the Horse Is Running
Let’s assemble the facts:
AMD waited five months to introduce a new speed bump for Athlons. Actually didn’t take them that long to start making them, the AXIAs started being made late January.
These chips are furnaces running at close to 1500Mhz. You don’t release chips like this if you have a better alternative ready. I don’t see how AMD could release anything more than 1400Mhz with this core to the public, and even that would be very dicey.
The initial Palominos are not being aimed at the high Mhz area, but the low ones, with relatively small immediate markets (for one reason or another).
AMD dramatically changed its roadmap only two weeks after they showed it to leading investors and investment firms.
What Is Happening?
I don’t know exactly what happened, and I suspect the only way to find out would involve hostage-taking and much torture. 🙂
However, I think I can give you a range of what’s happening, which is good enough for our purposes:
Best Case Scenario
AMD did have problems and delays with reducing power requirements.
They have now fixed them.
Delays in the various speeds of Palomino are primarily due to conversion of manufacturing facilities to the new product.
Worst Case Scenario
AMD had problems and delays with reducing power requirements.
They have not yet come up with ideal solutions, especially for higher-speed CPUs.
They’ve pushed off producing the more challenging processors hoping they can come up with one.
Were They Lying To Us Or To Themselves?
It’s not nice to say one thing at an investor conference, then say something quite different two weeks later.
1) AMD executives knowingly or semi-knowingly presented a picture to that investment conference that they knew wasn’t going to fly, or was subject to rather immediate change or
2) Accurate information was not given to the people making the presentation.
I can make just about as good a case either way on this.
The c’t roadmap indicates to me that roughly at the same time this presentation occurred, somebody at AMD knew changes were afoot. If the generals didn’t know, at least one colonel did.
However, the c’t roadmap doesn’t exactly follow the current AMD roadmap, which says to me that a final decision probably wasn’t reached until after it was made up. Quite possible there were several roadmaps made up for an executive meeting, and the executive chose a different option.
What is really boils down to is “Does the President and Chief Operation Officer of AMD know what’s going on in the business he’s supposed to be running?”
If he does, then he presented items at that meeting to people investing real money in his company knowing they were likely to change, and negatively, from the perspective of his audience. Sure, he had the figleaf of being able to say a final decision had not yet been made, which is any investor shouldn’t forget any future time Mr. de Ruiz ever opens up his mouth.
If he doesn’t, then not only should investors have a real problem with whatever comes out of Mr. de Ruiz’s mouth, he should have a big problem with his staff.