![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
. Who knows, maybe I'll jump to 939 pin.Originally posted by pakotlar
Now I'm more interested in seeing Tejas core. A64 isn't a justifiable jump to me because I'm addicted to multi-tasking. Although I think getting my hands on a 2.6ghz a64 would be spicy
edit: and prescott is def not at all interesting to me. what really miffs me is that intel could've released the same chip, called it prescott, without 11 additional pipeline stages. It would've smoked the a64, and I bet with a little work would've clocked to 4.0ghz. 3.6 ghz isn't out of range for the northwood cores even.
How? I don't see anything in the various benchmarks that would indicate significantly superior performance in F@H clock-per-clock over Northwood. Unless there is something in the architectural changes that would aid folding algorithms I'm not aware of...??
3.6GHz is not out of range for the Northwood core, but as clock speed increases, the yields of a given die process decrease. Since the .13µ process is at its EOL, ramping much beyond the 3.4GHz would have yielded Intel diminishing returns in profit. What you suggested about releasing a Prescott core identical to the actual chip minus the 11 additional stages sounds interesting, and may have been a fantastic performer, but doesn't take into account the very reason for these additional stages - architecture longevity. There are reasons we don't see 3GHz+ Pentium Ms and some of them are not marketing.
Originally posted by APOLLO
3.6GHz is not out of range for the Northwood core, but as clock speed increases, the yields of a given die process decrease. Since the .13µ process is at its EOL, ramping much beyond the 3.4GHz would have yielded Intel diminishing returns in profit. What you suggested about releasing a Prescott core identical to the actual chip minus the 11 additional stages sounds interesting, and may have been a fantastic performer, but doesn't take into account the very reason for these additional stages - architecture longevity. There are reasons we don't see 3GHz+ Pentium Ms and some of them are not marketing.
In order to compete against AMD's A64 with its advanced memory sub-system, Intel must design a chip that yields increasing clock speeds at a faster rate than its competition can bring its products to market. Witness the slowdown with P4 speed ramps in the last several quarters. The Netburst architecture was getting dated and AMD caught up in performance with the A64. Intel needed to regain the momentum and one of the best ways to do it was increasing the number of stages (as strange as that may sound). AMD may have won a small victory with the initial release of Prescott, but wait and see what happens by 2H of this year.
Prescott should quickly scale well beyond 4GHz while AMD will be stuck around the ~3GHz mark at the same juncture in time, perhaps a bit higher. The two companies are betting on completely different (m)architectural strategies to keep their market shares. I think Intel has won this hand, but we won't see it manifest for perhaps another 6 months, give or take a quarter.