What specific portions of IPP benefit most from Itanium2 processors?
That is, I would suppose that almost all IPP library methods are faster on a moderate (1.40GHz with 1.5MB L3 cache) Itanium2 than on the latest generation of Xeon/Pentium family of processors; but I suspect that while some are only marginally faster, others are remarkably faster (then again, my seat-of-the-pants impressions are often wrong) .
Are there exhaustive test results available for comparison? If not, what is the canonical theology (i.e., about specific functionality sets, rather than a general statement that Itanium2 is faster)?
Whatever the current state, how (specifically) do you expect the comparison to trend as all Intel processor lines mature?
 And indeed, I would generally expect a response curve where the size of the data being processed becomes a primary driver for the comparison (e.g., compared to cache size, tiny-data / medium-data / large-data / very-large-and-therefore-necessarily-streaming-data would exhibit very different IPP effectiveness ratios across the different processors).