Unless you can cough up a link to some benchmarks comparing a Northwood P4 to one of these new Celeron Ds, you don't have the right to continually make comparisons saying that they have equal performance clock for clock. Everyone knows that the Prescott P4 was slower than the Northwood P4 at the same clock speed (barring applications with SSE3 optimizations), due to the fact that Prescott had a longer pipeline and a higher latency L2 cache. A doubling of the L1 cache and L2 caches still didn't overcome these deficits on Prescott.
If anything, I would think that the new Celeron Ds are comparable to the Prescott P4, perhaps just a little bit behind it on a clock for clock comparison. It depends on where the "fine line" is on L2 cache size being just enough and too little. As we witnessed from the Northwood Celeron to Prescott Celeron, a doubling of the L1 and L2 cache had a MUCH bigger effect on performance (a 10% increase clock for clock) than the Northwood P4 to Prescott P4, where virtually no improvement, and actually some slight decreases in performance were noticed. That was due to the fact that 128 KB of L2 cache just isn't enough on the Netburst architecture to prevent pipeline stalling and flushing. The 256KB L2 on the Prescott Celeron seemed to be the Messiah, greatly increasing performance despite that fact that it still had the same slow latency L2 and increased pipeline length of its big brother P4. Still, comparisons of the Prescott P4 to the Prescott Celeron showed that the diminished L2 cache of the Celeron D still hampered its performance. The question now is whether the 512KB L2 cache will show similar gains in performance that were shown in the move from Celeron to Celeron D, and whether this gain in performance will bring the new 65 nm Celeron D up to the same level of performance as the 65 nm Pentium 4s (which, by the way, still use the same pipeline length and slow cache latency as the Prescott P4, explaining why we don't see any real improvement with the 65 nm P4 over the 90 nm P4 in terms of performance on a clock for clock comparison).
According to the Extreme Systems link you posted (which I gave to you), a 5.0 GHz 65nm Celeron D games as well as a 4.5 GHz P4, so there is obviously a noticeable performance hit going from 1MB L2 to 512KB L2, meaning that the new Celeron Ds SHOULD NOT EQUAL THE PERFORMANCE OF A NORTHWOOD P4 ON A CLOCK FOR CLOCK COMPARISON. Saying different is just ignorance. :?
Still, the new 65nm Celeron Ds are a great processor, without question the best single core Intel processor yet produced. If it can easily overclock to 5 GHz on air, on stock voltage, using the stock Intel hsf, staying under 50C, consuming less than 130 watts, and then go toe to toe with a 3 GHz Opteron and come out the winner, then it definately has my vote.
