Originally posted by: Martimus
I am not disappointed at all. The IPC improvements are actually much better than I expected. It is expensive, but so too was the GTX280 when it was released. It will come down in price if Deneb is competitive.
+1 :thumbsup:
OP also consider that at the time of C2D release you were likely comparing the performance and performance/watt of either a 130nm Northwood or a 90nm Prescott or 90nm Athlon X2 to the performance and performance/watt of 65nm C2D.
The node shrinks alone afforded the C2D team to use massively more (in number) transistors on the 65nm C2D design compared to the number of transistors the 90nm P4 and X2 had at their disposal.
So you had the perfect setup for creating product differentiation - a node shrink meant more transistors (more performance) and it also meant lower power (more performance/watt).
130nm Northwood had 55million xtors
90nm Prescott had 125million xtors
90nm Athlon 64 X2 (2x512KB) had 154million xtors
65nm Core 2 had 295million xtors
So you give the C2D CPU designers twice the number of xtors as you give the Prescott or Athlon X2 design teams and you kinda expect them to deliver some higher performance numbers.
You also give them 65nm node transistors instead of 90nm node transistors...and you kinda expect things to turn out well for the performance/watt metrics.
This is what set the stage for your experience and nostalgia of 2006 C2D release.
Now look at what knew about Nehalem before it was released. Yorkfield was 45nm already, already had those low-power transistors built into the performance/watt. No new node transition for Nehalem.
Also Yorkfield was designed with a transistor budget of 810million transistors (two 410million Penryn MCM'ed together). Nehalem's transistor budget actually shrunk to 731million but now includes the transistors needed for all the uncore stuff (QPI, mem controller, PCU, etc).
So you knew that C2D blew your socks off because it had the advantage of an entire node shrink (power reduction) and twice the transistor budget (performance) and you also knew that Nehalem had neither of these advantages over Penryn (same node, smaller transistor budget)...and yet you claim your expectations were realistic and you feel rightfully disappointed in the Nehalem results?
Pardon me while I shake my head and mutter "what a world, what a world" for a few seconds.
Personally I find the Nehalem results to be simply astounding. To squeeze as much as they did out of the existing 45nm node, and doing it with fewer transistors, that's rare. Very rare. You'd be hard-pressed to find me an example of this ever happening on another occasion in the history of the desktop PC.