This extraction is not quite fair, nor correct. The entire memory sub-system for Nehalem has been reworked, you cannot look at the off-die DDR2-DDR3 comparision and make the assumption you are making.
In terms of marvels, who knows ... we have very little info to go by ... however, the hype machine of Geslinger compared the leap of Nehalem over Conroe to the magnitude of Conroe over Netburst, this indeed would be marvelous. My take on it is that it will be less overwhelming in single thread, but multithreaded situations will be impressive on this magnitude... Anand's preview data seems to indicate that this is indeed the potential... 30-50% gains depending on the app (multithreaded of course).
This will be even more impressive in server, where the mem BW issue is much much more pronounced... Kanter (RWT) seems to think Nehalem will be nothing short of a miracle. Time will tell.... however, initial indications are very positive that your downplay of Nehalem is probably not warranted.
Jack






Reply With Quote

Bookmarks