When the first generation of the Pentium 4 was released, the processor was based on the 0.18 micron Willamette core. With 256K of L2 cache, and running on a 400 MHz (100 MHz quad-pumped) FSB, the first generation of Pentium 4 CPUs were outperformed (on average) by a similar-model Athlon XP, and showed relatively low overclocking potential. Not being one to just sit idly while a competitor tears apart their processors, Intel released a much higher performance 0.13-micron Northwood core, complete with 512K of L2 cache and an upgraded 533 MHz FSB. Due to the higher clock and bus speeds, along with the larger L2 cache, the Pentium 4 Northwood faired much better against the Athlon XP, and its smaller core was very well received by the overclocking community.
Shortly thereafter, Intel resurrected their value line of Celeron processors, which had previously used the aging Socket 370 platform. Intel released Celeron 1.7 GHz and 1.8 GHz models, which are based on the Willamette Pentium 4 core. Intel didn't move immediately to the Northwood core, and to properly differentiate between the two processor line, the Celeron had its L2 cache chopped in half to only 128K, or one-quarter the level of a Northwood Pentium 4. The original Socket 478 Celeron processors shared the same problems as the original Pentium 4, and were difficult to overclock, used the 400 MHz front-side bus, and ran hotter than the newer Northwood. As such, the first generation of Celerons was not well received.
Let's fast forward six months to the release of the 2.0 GHz Celeron processor. The Celeron has now been upgraded to the 0.13-micron Northwood core, thus allowing it to run at much higher speeds and cooler than before. However, even with a full 512K L2 cache in the Northwood design, Intel stuck with the 128K cache for the Celeron 2.0 GHz and the new Celeron remained stuck on the 400 MHz bus. The performance of the Celeron 1.7 GHz and 1.8 GHz was adversely affected by the small amount of L2 cache and 400 MHz FSB, but Intel has not done anything to alleviate this. However, we are looking at a full 2.0 GHz of clock speed, and the 0.13-micron core should allow much higher overclock speeds than was previously possible. Could this be the old Celeron 300A reincarnated for 2003?
By far the most exciting news to come out of this review is the 2.0GHz Celeron's overclocking potential. On our Asus P4S533 reference system we had no trouble running the Celeron at speeds up to 3.0 GHz, and even a hair higher. Because Intel locks the multipliers on all their chips, we had to resort to FSB overclocking to achieve the overclock that we did. Because the 2.0GHz chip's multiplier is locked at 20.0, this meant pushing the FSB up to 150 MHz (600 MHz quad-pumped) which we did not have any problem doing.
Despite using some basic air cooling we encountered no problems whatsoever at 3.0 GHz. In fact, we didn't even need to raise the core voltage past its default operating specification. All of the benchmarks were run with a single crash or hiccup. However, above 3.0 GHz we started having trouble, and without an increase in cooling power or voltage it was not possible to sustain an acceptable level of stability at higher clock speeds.