Intel's new Pentium 4 chipsets

Nebojsa Novakovic is an experienced journalist who deals on a day-to-day basis with real world configurations for PCs and other systems working in enterprises.

This is his inaugural test of the latest Pentium 4 chipsets, viewed, not from the consumer or end user point of view, but in the light of practical configurations system builders might make and sell.

He is at the Computex show in Taiwan all next week and can be contacted either at this email address, and is the INQUIRER's regional editor in the Asia Pacific area.

WHILE CPUS are obviously its bread-and-extra-creamy butter, Intel has come back into the chipset arena with a vengeance.

The newly updated Pentium 4 chipset offerings, the i850E, i845E and i845G, boast not just faster 533 MHz FSBs for the new Pentium 4 CPUs, but also improved memory support (a lot of it "non-validated" yet, though) and new I/O based on the ICH4 south bridge -- again "non-validated" for i850E, but guess what, most i850E mainboards will have ICH4.

How much benefit is there from a faster FSB if your main memory speed still lags behind heavily? Should a system integrator bother spending extra time and resources for scout for PC-3200 (DDR400) memory for his new PC line, and will it bring a real performance advantage? Will PC-3200 be faster than PC-800 RDRAM dual channel, and what about still expensive PC-1066?

For this the first of many INQlab tests, we set up four nice systems, with i850E, i845G and i845E chipse, plus SiS645DX.

I actually bothered to buy two high-speed DIMMs from Corsair (one PC3200 CL2.5, another PC3000 CL2 - declared ratings). The idea was this: let's see how do these compare with identical CPU and memory settings, as well as how far will they go in the memory speed above the usual specs?

Since this is aimed at system integrators who are supposed to provide warranty to their users, the CPU or FSB overclocking was out of the question: the whole thing must be a performance-maximisation text within the guaranteed CPU, board and memory specs!

What we got were four varieties of one identical configuration: a Pentium4 2.4 GHz PC with 512 MB RAM, 40 GB 7,200 rpm ATA100 HD, ELSA Gloria 4 (Nvidia Quadro 4 900XGL), running Windows XP with Nvidia 28.32 drivers - nothing unusual.

Now, besides the chipset varieties, we also induced the FSB choice of 400 or 533 MHz on two of the configurations, to show the FSB change effects if the memory speed stays the same. We ran a multitude of the usual benchmarks (Sandra, CLiBench, PCMark, Sysmark 2002), and also looked at the stability of the platform in terms of sales trustworthiness (i.e. do you dare to actually integrate, guarantee and sell such a config to anyone).

These are our results.

Test Results