Navigate: Advice Beginners BIOS Guide CPUs Links Mainboards Memory Network Storage Video/Sound Cards Contact Forum SiteMap Sponsors WebNews Home	.	.
Prices: Mainboards ABIT ASUS Chaintech Shuttle Soyo Tyan CPU Intel P4 2.4C-800 P4 2.6C-800 P4 2.8C-800 P4 3.0-800 P4 3.2-800 AMD AthlonXP XP 1700+ XP 2000+ XP 2400+ XP 2500+ XP 2700+ XP 3000+ XP 3200+ Athlon64 Athlon64 3200+ Athlon64 FX-51 Opteron Opteron 240 Opteron 242 Opteron 244 Opteron 246 Memory Corsair Crucial Kingston Mushkin OCZ Search Prices:

Time to wrap up this review, there are some 200 more data points that we could post but essentially, they would all show the same, that is, the Pentium4 2.4B is the second fastest CPU on the market, surpassed only by the one that came after it. Unlike in Hitchhikers guide to the galaxy, this one has a name, Pentium4 2.53, easy to remember but rather uninspiring (unlike Penteron). So what's the real conclusion here? Four hundred and fifty fps in Quake3 Arena beat the 370 fps we got with the overclocked Athlon XP 2100+ on the Shuttle AK35GT2/R? One hundred and eighty five fps in Expendable (XP2100+) blow away the Pentium4 along with all its Rambus. Content Creation Winstone 2002 vs. 2001, have the winds of trade changed in favor of Intel?

With all love to AMD, there is no doubt that the new P4 using the 533 MHz FSB is the fastest x86 processor but endearment and price have the same root, at least in Latin language and so it is not too surprising that the price tag associated with the P4 is about twice as hefty as what one has to shell out for AMD's flagship. But then, the die size is bigger, meaning, you get more CPU real estate for the money....

Reviews are not about humor, though, it is hardcore business that we are talking about and that we are influencing by writing reviews. This said, it is important to look over the results and try to make some sense out of them.

What our results show is that at stock settings, Rambus 800 still holds a slight margin over the DDR systems. There are several cautionary statements that need to be made, for once, Intel boards are always slower than whatever third party manufacturers are capable or daring of cooking up. On the other hand, this may be negated by the fact that the SIS chipset board used here was not really intended for a 133 MHz FSB CPU and this could negatively affect the performance numbers.

On paper and in synthetic benchmarks, Rambus certainly looks great, the only gripe we have with this is that for the performance oriented user, DDR offers a huge headroom for overclocking. Knowing that PC1066 RDRAM is still hard to come by, it does not appear as if those RIMMs would withstand a terrible amount of overclocking but then, we could be wrong and the RIMMs might go all the way up to 150 MHz or beyond.

Playing Devil's Advocate, we say that they won't, not to mention the technical difficulties associated with a high speed bus that made it necessary to substitute higher grade clock generators just to get the 1066 MHz interface stable (if workable at all) on the next generation of PC1066-compatible mainboards. Granted that there is always a possibility of making things work, the question is still at what cost and there it is where, on the board level, on the chipset level, on the component level and last not least on the memory level the equation looks much in favor of DDR.

On the other hand, there is the fact that, aside from "inflated" gains in synthetic memory benchmarks like SiSoft Sandra (no offense intended here), the real life performance increase through the move to 133 / 533 MHz FSB in conjunction with the use of PC1066 RDRAM does provide some more than healthy performance boost for the i850E platform. Up to 16% performance increase in Quake3 Arena based on nothing but lower multiplier and higher bus speed are telling a very convincing story about what is wrong with the GigaHertz race, at least in its present implementation based on ever increasing multiplier values.

On yet another hand, there is the issue of increased chipset latencies that become necessary because of the higher FSB. Where, at 100 MHz, a zero bus turnaround may have been possible, the 133 MHz FSB will require additional wait states. The same may or may not apply to the dynamic bus inversion with the difference that instead of 1 latency cycle we may now look at two. Something along these lines must be what is going on with the move from the 100 to the 133 MHz FSB, otherwise, there would be no logical explanation for either the unbuffered SiSoft Sandra memory scores or the Expendable scores. Particularly the latter almost lead the entire elaborate scheme about benchmarks and the theory behind them as well as their relevance for practical purposes ad absurdum. And, no, it is not ironic.

After all these theoretical considerations it is about time to look at what would be the highest performance we can expect from the Pentium4. What I am getting at is the fact that we were using an Intel board bar any overclocking features which naturally puts the RDRAM system at a disadvantage. Still, there does not appear to be that much headroom in RDRAM as in DDR, however, as mentioned earlier, we were in no position to challenge the potential of RDRAM.

One issue I did not mention yet is relating to the overclocking of the new speedgrades of P4. That is, the lower multipliers, of course, also translate into diminishing gains for overall clock speed increments, meaning that we are once again, running into the issues of PCI speed etc. and the associated risk of corrupting the HDD or malfunctioning PCI devices. A difference between an 18x multiplier and a 24 x multiplier does not appear to be that dramatic but what goes around, comes around. That is, all the little advantages listed above for lower multiplier values backfire at the point where a 1 MHz increment in FSB only yields 18 MHz instead of 24 MHz. For overall clock speed, 5 MHz FSB overclocking result in a difference of 30 MHz and 10 Mhz are equivalent to a 60 MHz difference based on either 18 or 24 x multipliers. Nothing too significant but there are still people who still get the kicks out of raw clock speed, mind me saying that some within Intel's marketing department may fall into this category?

Rants aside, one thing is very clear, that is, in the first few months of this calendar year, the Pentium4 has undergone a dramatic metamorphosis from the ugly duckling to be continued ...

next page:    => more =>