Dothan Over Netburst: Is The Pentium 4 A Dead End?

Intel - From Odyssey To Course Correction

Imagine an old product that officially disappeared from the marketplace several years ago returning from the grave to outperform an innovative new product that was predicted to have a rosy future. Intel stubbornly insisted on evangelizing the advantages of the Pentium 4 design in countless advertising campaigns, aggressively pushing it as the architecture of the future. But what does the future hold for this supposedly innovative and forward-looking design? Apparently, not a great deal.

It seems that the much-lauded Netburst architecture is about to be put out to pasture. At least, that's what Pat Gelsinger, GM and VP of Intel's Digital Enterprise Group, recently confirmed in an interview. Of course, this move has a great number of specialists and insiders scratching their heads, wondering how it could be that Intel, the market leader, suddenly feels the need to start backpedaling.

As early as 2000, THG observed that the Pentium 4's performance was clearly inferior to that of its predecessor, the Pentium III, on a clock-for-clock basis. On the other hand, the newer design had one undeniable advantage: it could reach much higher clock speeds. This was the weakness of the Pentium III design, which hit a clock speed barrier as a result of thermal problems - a fact that THG helped uncover. More information about the Pentium III's thermal problems can be found in this article Intel Admits Problems With Pentium III 1.13 GHz: Production and Shipments Halted .

At the beginning of the odyssey that was the Netburst architecture, Intel clearly spoke out in favor of the serial Rambus memory technology as the future memory standard. However, the high cost of the Rambus design led to its being displaced by the cheaper DDR memory standard. Intel's next step was to split the P4 processor into two virtual units, a technique that became known as "Hyper-Threading" or HT. In our article Single CPU in Dual Operation: P4 3.06 GHz with Hyper-Threading Technology and accompanying video, we were able to demonstrate that multimedia applications could indeed be accelerated using HT.

At this point, the Netburst architecture had already reached its zenith, as the following chip generations showed diminishing performance returns despite their ever-increasing clock speeds. At the same time, the thermal dissipation loss reached astronomical levels with the introduction of the 90 nanometer P4, codenamed Prescott. Even today, Intel has yet to give us a satisfactory reason why the Prescott P4, with its 90 nanometer design, has a higher thermal loss than its 130 nanometer predecessor, the P4 Northwood, when comparing chips at the same clock speed.

Even more interesting is the fact that Intel never really killed off the Pentium III. Large parts of the highly effective and much less power hungry architecture found its way into the very popular Pentium M notebook processor. Indeed, when analyzing the architecture of the Pentium M, it becomes clear that it is much more similar to the Pentium III than the Pentium 4!

As a result of the miniaturization process from 180 nanometers (problem child Pentium III 1.13 GHz) to 90 nanometers (today's Pentium M Dothan) and the optimizations subsequently made to the die and the silicon itself, the Pentium M now proves to be the better processor overall. The processor generates far less heat and can process more instructions per clock cycle, making it much more efficient.