While we're waiting on Intel to ship something competitive in the present, it sure is fun to listen to Intel talk about what it's going to do in the future, isn't it? Intel must surely think so, as this has been the company's basic marketing strategy for the last six years...;) But every year "the future" gets pushed a bit further out, unfortunately. WIll this be the year the future arrives? Guess we'll have to wait a little longer to find out...shucks...

...And I can remember a time when Intel didn't even announce a product until it was already in the channel and widely available. Things have certainly changed.

that's incredibly aggressive...does Intel really think this is feasable?

#24, Yea, I don't know what their engineers were thinking. They went all the way from 1.3Ghz up to 2Ghz with Socket 423. I would imagine it's one of the things that Intel engineers are probably embarrased about. How long did Socket 423 last until Intel came out with 478?

If you have a look at the Willamette processor (like, for example, here: http://techreport.com/reviews/2001q2/pentium4-1.7/index.x?pg=1), you can see that it's using a transposer (IIRC, haven't read about it in years, the green thing, anyways) that interfaces the PCB carrying the die (brown) with the PGA socket. This brown piece of PCG does use a µ BGA interface, as you can see from the bottom-up shot.

My guess is that the problem was somewhere in the µPGA pin or socket area, maybe the production not being up to scratch or lacking sufficient volumes. Had µPGA technology been feasable in 2000, Willamette would probably have launched w/ a µPGA Socket 423. (think actual S423 Willamette - transposer + pins instead of solder balls on the brown PCB)

Edit: Sorry, meant to reply to #24!

I don't get where are this "Intel is changing sockets too often" thing is coming from. On the desktop they've had, count `em, three different sockets since the P4 came out. Socket 423, Socket 478, and Socket T. Each change was made for technical reasons and not seemingly for anything to do with marketing or to prevent interchange of parts.

Conversely, on the AMD side we had one socket for the Athlon/Athlon XP and to totally confuse matters three different concurrent sockets for the Athlon 64 and Opteron processors. And because the memory controller is integrated into the CPU we must change the socket to protect people from themselves and we add one more called AM2.

I'm not even getting into the mobile or server side of things.

I'd say that the Intel side is far less confusing and has gone through fewer changes at least for desktop computers. It really isn't in their best interest to keep changing sockets.

For a company with the resources of Intel, running two independent design teams on overlapped schedules is doable (albeit expensive) -- essentially they were doing that already, with the Pentium M team running independent of the netburst team (not to mention the Itanium guys off in their own little universe) -- and the netburst team no longer has anything to do. Essentially Intel is still doing the same thing, but now both teams are working on a complete family (server, desktop, and mobile). And introducing process improvements on an offset schedule makes total sense: you want to use a proven design with a new process, and you want to use a working process to fab a new design.

In the past Intel's design teams have been rather competitive and exclusionary, but if they're going to stick to this kind of schedule they need to exchange a lot of information about what is working and what isn't during the development process so that each group can build on the other's accomplishments. The real risk for Intel is that if one of the teams paints itself into a corner, a two year timeframe just doesn't have the slack to scrap a design and start over. Of course they can stretch an existing design to cover a longer timeframe -- and given corporate adoption rates, a longer timeframe might actually make sense -- but now that they've made this claim any slippage beyond two years is going to be seen as failing to deliver. However, every time they do successfully introduce a new design they get a lot of easy PR.

This is proabably a shot at AMD as much as an actual statement of intent: this is another way Intel can better leverage its resources against AMD, who can't maintain that kind of schedule. Of course, it's not clear that they have to: they just have to compete on the basis actual product, not on the pace design introductions.

Unfortunately changing the basic memory type means a change so large that you should change sockets to prevent the interchange of parts. Thus the AM2.

Intel does this for the same reasons, however Intel has been constantly tinkering with the P4 trying to keep up with AMD, resulting in a lot of changes, and new sockets.

Neither company changes sockets for anything other than compelling reasons.

oh great...so we will get a new socket every 3 months instead of ever year like they currently do. last time i looked at intel it was pretty scary how many actual chipset/sockets/chip combinations they had on the market....

me ill take my socket 939...unplug single core...plug in dual core...all good. not that amd only have a single socket (although thats the rumour with AM2 isnt it?) but at least it has been somewhat less than intels mess.

I wonder how much better the cores will be, rather than just very radical different from one another. But it sure is a whole new approach tho.

"This is a radically new strategy for Intel considering it has ridden the Pentium 4's Netburst architecture for five and a half years"

and good riddens it was!