Thanks to all who responded to my question:
This is what you told me:
1) Most XPs sold lately were made between weeks 20 to 26. The latest one reported was week 32.
2) There were earlier ones floating around, but the most extreme example of this we saw were a couple reports of 1900+s made at the end of 2001 being recently sold by a U.S. reseller.
3) If I had to breakdown the source of these chips, it would have to be 1) Newegg and 2) Rest of Planet. Obviously skewed, but it proved to be a very useful skew (see below).
According to AMD, Dresden wafer starts (which were more or less 5,000 a week for this whole period) were entirely or almost entirely Palomino until the end of 2001.
During the first quarter 2002, that 5,000 Palomino wafers slid to about half that by the end of the quarter.
During the second quarter, that 3,000 slid to effectively zero at the end of that quarter.
The pattern of weeks reported is consistent with the wafer starts. Week 32 is probably the last or close to last group of Palominos made, and there should be relatively few of those.
So weeks 20-26 Palominos are at the tail end of Palomino production. It should also be noted that the Newegg CPUs tended to have later weeks than the rest.
Since the Newegg website has periodically run out of processors and ordered new shipments, their results are probably a good indicator of inventory backlogs up the line. Newegg is also one of the few places that are offering low-speed TBreds.
While we can see the light at the end of the tunnel, we’re not too sure how long that tunnel is. It sure doesn’t look like AMD will run out of Palominos anytime soon.
The New Duron: TBredA?
Remember, Durons are about to become history, and XPs are supposed to replace them. From AMD’s pricing perspective, XPs already are at Duron-like price points.
Let’s pass on the subject of how long Durons may linger in inventory given the history of its elder brother.
At some point though, AMD has to run out of the things, and since the XP line will become the Duron line, the Palominos will get replaced by TBreds.
But which one? TBredA or TBredB?
I think they’re going to be TBredAs, and I’ll explain why.
There’s a physical difference between TBredAs and Bs. The Bs have an extra metal layer (and another 400K transistors). They cost more to make. They may require other changes to manufacturing process.
When you’re hurting for money and out to save every cent, why spend more money on the manufacture of low-speed processors that don’t need the improvement faster? It doesn’t help the average user. The only people it helps is . . . uhhhh . . . us, and the more money left in our pockets, the less there is in AMD’s.
There’s already some real evidence of this. The very early 1700+ sold by Newegg were made around week 32, and had old codes like AIRGA. I’m using one right now (and it’s not doing well at all).
However, it looks like the 1700+ CPUs (and the 1800+ TBreds) are now chips made very recently with codes like AIUGA. This means AMD is still actively making TBredAs. That makes them do better that the older chips, but not nearly as well as TBredBs, which keeps the speed demons spending $180 for 2400+ rather than the $70 they’d spend with a 1700+ TBredB.
Please note something:
I would venture the guess that the two are now being made the same way outside of that extra metal layer and extra transistors, and that the difference in overclockability is primarily due to that extra metal layer and extra transistors.
While the newer processes make the TBredAs do better that the older versions, they don’t do nearly as well as the physically modified TBredBs. For AMD, this is good, because it keeps the speed demons spending $180 for 2400+ rather than the $70 they’d spend with a 1700+ TBredB.
It’s hardly realistic or reasonable to expect AMD to spend more money to make a CPU just so they can get less money from us. If you think otherwise, don’t write me. Get a taste of the real world and shovel it to AMD. See what kind of answer you get. 🙂
I would also venture that any TBred replacements for Palominos anytime soon will most likely be TBredAs. It’s possible that a few TBredBs might sneak in, but they’re likely to be rejects unable to sustain high speeds.
Eventually, I think you’ll see TBredBs migrate downward, but only when doing that is to AMD’s advantage. I think the first sign of that will be seeing TBredBs showing up at the 2200+ level (outside of selling marginal chips).
I don’t think you’ll see TBredBs migrating to lower speeds until it is to AMD’s advantage to do so. In all likelihood, that will be when a) AMD needs to free up some capacity to make Clawhammer and/or b) when UMC starts making processors for AMD.
I still think you’ll see an 1800+ TbredB. Problem is, that may not be until next spring or summer.
Again, thanks to all who responded.