Wingznut PEZ, look at my .sig. Supercomputers run 64 bit. The benchmarks I have seen for Itanium come from that. They are slower than the corresponding SGI/IBM/Cray scores. Now is that because it is slow, or the compilers, or the MTH that SGI is using, or is it because they were the early cA2 chips, or who knows what? I do not know what the reason is, nor do I care as that is not really my department. All I care about is performance.
And so what if Itanium is late and slow. It was Intel's first attempt. In my philosophy, failure is an option. After all, the Wright Brothers crashed before they flew, and the first airplane death was with them at the helm.
I bet Intel learns from the Itanium failure and produces a good McKinley. If so, then Itanium was not a failure. I have not seen any benchmarks on McKinley, yet. I have heard good things about it though. Too bad my friends at KAI won't tell me anything about it.