McInerney was coy in and interview:
"I don't want to go too far out in time, but you've
had people talk about 'Can you come up with one architecture
that runs both software stacks? Can you use binary translation
and so forth?' We continue to play with those ideas, but in the
end we still see having a dedicate Itanium core that's optimized
to Itanium still gives us the best solution."
At the moment, the processor roadmap (see chart below
) has the Xeon family moving from
its current E7 to Ivybridge in 2013 and to Haswell at some point after
that. Itanium, on the other hand, has the 9500 (code-named Poulson),
detailed this week, then Kittson at some point the future. Then.....?
What's after Kittson?
"Itanium's on a two- to three-year timeline. We don't spin them out
like workstation products," McInerney said. "We've shown a roadmap
that takes us out to mid-decade. A post-Kittson product is
.... a second-half-of-decade product. We don't even disclose
Xeon out that far."
It's hard to imagine Itanium fading over time. There are a lot of
ancient architectures still chugging away in warehouses and closets
around the world. But server economics get less expensive by the
year, and technology adoption habits change with it.
McInerney would never say never, even if you could egg him along
"Microarchitects are always innovating. It's same
with the architecture teams. They know the problems they need to
plugs eight-core Itanium into five
packs eight cores into Itanium CPU
ends work on Linux for Itanium