Tony, Re: "Sanders said very clearly in answer to a question something like "so you're in production on 130 nanometer...?" that AMD is not in production on 130 nanometer yet, that they were still in the qualification phase, and that first silicon was coming up this quarter. Sanders said first Silicon was not achieved yet!
So what does that mean re first shipments, intro, volumes? Sounds like a big slip to me."
Let's just assume for the moment that Jerry used the wrong lingo (oops), and that he really meant to say that the first production wafer starts had not returned from the fab yet (first silicon meaning the first production wafers). In this case, I can see a sizable .13u wafer start as soon as they receive back samples that they consider to be production worthy. If they get the production silicon they are waiting for within the next few days, they could begin a high priority rush through the fab, and get out low volume production shipments before January.
Let's consider that the optimistic case. Next is the realistic case.
Let's assume for a moment that AMD is still in the sample phase, but one or more factors are gating production. This might include bad yields, a last minute bug, poor analog or thermals, defect issues, a failure with the focus, position, or quality of the equipment, or any number of things. Let's say that Sanders considers first silicon as the first Thoroughbred wafers out of the fab that meet the standards of something that is shippable. Since real life issues have no definite timeline, you really can't tell when they will get things right. But let's assume that they think they are close, and solve the issue in a couple of weeks. They will then put a high priority wafer lot through the fab, and get their "first silicon" by the beginning of next year. At this point, either they will have more problems, or their fixes were successful, and they can begin high volume production. In this case, estimated launch data would typically run from late February to March, or possibly later if they can't fix their issues.
Lastly, let's consider the possibility that Sanders really did mean first silicon, and that Thoroughbred has only just taped out, but hasn't been received back. If Tbred is a simple shrink of Palomino, then production will be about 4-6 months from tapeout, which may suggest an April or May launch. That's if no bugs are found on their first silicon. If there are, it might get pushed out to June or July, depending on how many steppings they need before they fix the problems.
Personally, I don't tend to go for the final theory, simply because my impression is that AMD is far more confident than would imply that late of a release. The first (optimistic) theory is less appealing, too, since AMD seems to be putting a few more Palomino speed grades on their roadmap, and it wouldn't make sense for them to do so if they had Thoroughbred ready as early as late December.
That brings us to theory #2, which suggests a February or March release, and that almost corresponds to AMD's roadmaps. I think that they will try to make a mobile release, first, in order to gain market share in higher margin markets, and then they will launch TBred for desktop in mid-Q2. If this is the actual launch date, then they will face a 2.4GHz Northwood in the same quarter, which has a good chance of being the performance leader until TBred arrives.
Rather than making any more second guesses, though, I'll just remain on the watch to see if any new information arrives. It appears that AMD will have to make Palomino work for longer than they may want, but some good news on the side of production could mean TBred a little early.
wanna_bmw |
