>>> I would like to know if anyone has made an in depth evaluation of >>> the difference between SSE3 & the Altivec implimentations in the >>> G4+ and the 970? >>> >>> This to me appears to be where Intel falls short. They appear to >>> do better in Interger computation, the reason that Mathematica >>> did so well, but fall short by a wide margin in the vector, the >>> strength of the PPC platform. > For me as a scientist, the AltiVec is always been useless because it only ran single precision, where all around me everyone uses double, if not quad precision. However, the AltiVec and its API looked beautiful, including the only prefetch-for-write in existence, for as far as I know. Victor.