I've never heard of the P3 core being migrated to 90nm, which this articles
says the Shelton is. The Xbox chips are still being produced on 130nm
(possibly 180nm) production lines. Since Intel has so many fabs, this is no
big deal for it to be running multiple process nodes.
Intel usually keeps two process nodes going for most of their fabs and
does a bit of leapfrogging, ie those old 180nm fabs would probably get
upgraded to the 90nm fabs while they keep the 130nm ones kicking
around until the next upgrade.
The XBox chips were being produced on the 180nm production line
initially, but producing them on a 90nm line would be more economical.
They are probably selling enough chips just in the XBox to make the
switch-over worthwhile, and if they could additionally get a
dirt-cheap processor for developing markets out of the deal it might
just seem worthwhile.
Just a though, I don't haven't even heard any rumors to back this up
or anything, just sort of makes sense in my mind, MUCH more so than
trying to put out a P4 or Pentium-M based Celeron with no cache. The
P4-based chip with no cache would be an absolutely abysmal performer
(probably about on-par with a PII-233) and would still be a pretty
darn big die (probably 60M+ transistors). A Pentium-M based Celeron
with no cache would perform better but it would probably require a
fair bit of tweaking to manufacture the thing cheaply. A 90nm version
of the XBox chip, on the other hand, would be DIRT-CHEAP (less than
10M transistors and a die size probably down in the 20mm^2 range) and
could straddle two markets, giving it enough production to be
meaningful. They could probably also sell some of these chips as part
of their line of embedded Celeron processors as well.