Intel Corp (INTC)

[chipguy]

In fact, I would argue that if Intel didn't have the Itanium distraction, its server roadmap wouldn't have been in shambles in the time period between 2003-2007, and AMD would not have had any success with Opteron.

Opteron's success had nothing to do with 64-bitness. Customers
bought it because it ran 32 bit x86 code better than Intel's P4
based server chips. The reason it did was partially because P4
uarch was botched at 90 nm but mostly because Opteron integrated
memory controller and interprocessor comms on chip while Intel
stuck with multi-drop FSB plus chipset based architectures. This
gave AMD better memory bandwidth and latency in the crucial 2
and 4 socket server markets.

None of this has anything to do with Itanium unless you're saying
all the brilliant designers at Intel were working on Merced and
McKinley and the B team was left on their own and came up with
the ultra speed-racer based P4 strategy which persists across 4
process generations. I don't buy this. There were some pretty
bright guys who worked on P4.

I can start expressing some of my true feelings about Itanium and how it has become such a millstone around the necks of anyone involved with it.

I am not surprised. The program came off the rails right off
the start with huge problems with Merced (mainly due to one
basic fatal mistake) and never really recovered. Itanium became
a very popular stick to beat Intel with by the tech press and both
RISC and x86 competitors because Intel was so overwhelmingly
dominant in PCs with x86. Hardly surprising that IPF projects
weren't a popular career path within Intel after 2002 or so.

The reality was after McKinley the resources Intel put into
Itanium were paltry compared to x86 and it was basically one
really late i2 shrink after another. But it was far more a public
relations millstone problem than a resource sink. Intel's main
problem with Itanium was it should have either fully resourced
it to go toe to toe with POWER (something like Poulson uarch
in 90 nm to beat POWER5) or just killed it outright instead of
dragging it out for another decade like a dead dog tied to the
bumper of a car.