I expect Intel/HP will continue with Itanium until IBM decides to go foundry for its silicon.
I don't understand your reasoning. If IBM falters with
its proprietary platforms then that increases the value
of IPF to Intel and HP - the last truly mission critical
man standing as it were.
I don't see why Intel would cancel IPF as long as HP's
non-x86 systems were selling sufficiently well as to
make its IPF MPU business profitable. HP undertook an
extremely painful 15 year effort to port its OSes, ISVs,
and customers, from 3 different RISCs to IPF. The idea
that HP would force another move from x86 to IPF for
short term performance/cost arbitrage between IPF and
Xeon products seems ludicrous. In about two years IPF
will go from 65 nm 4 core to 32 nm 12 core with Poulson.
Also keep in mind that Poulson will have the first brand
new IPF microarchitecture since McKinley was designed
in the late 90s with no customer application trace data!
A matchup between a 32 nm 12 core Xeon MP and a 32 nm
12 core IPF chip with a modern CPU core will be *very*
interesting to say the least.
However if Poulson doesn't appear until 2012 on 32nm it will still be a node behind x86 which should be full bore on 22nm by then.
Client silicon maybe but not a 22 nm Xeon MP. Especially if
Xeon MP is sticking with big die sizes like Beckton.
Market Data 
Markets 
AI
