The i875P chipset has strengthened the position of the most efficient solution after almost two months of its existence, and now it keeps on carrying the baton taken from the i850/E in detour of the iE7205. But its market position is not clear as compared to the previous chipsets developed for workstation/enthusiasts, and even the AGP Pro port, a traditional accessory of the workstation class which often came with iE7205 based boards, will be attached to i875P based boards only in 50% of cases. Many board makers grasped the idea of an expensive board fully compatible with "usual" products, the boards based on the Canterwood do not move to a separate niche - they just take upper positions in wide lines of respective companies. 

Sure, the price for a board with such desktop chipset (as if accelerated) is not low, and different companies justify it differently: some add rich suites of accessories, others install controllers of exotic peripherals... Such boards are sold mostly to home enthusiasts, and the marketing tricks are widely used here. Thus, almost everyone uses an expensive chip of the Intel's 1Gbit controller, because of the name rather than because of the characteristics. The only unique technology of the i875P - PAT - is also taken advantage of, though ASUS is pressed from above as it tries to use this term out of the frames of the context confirmed. 

Among all players in this sector it was only EPoX who tried to fix a reasonable price. Its 4PCA3+ looks poorer than the other boards, but the price is also more attractive. However, the EPoX's board won't be tested today. This time we have mainboards from such companies as ABIT, ASUS, Chaintech, Gigabyte, Intel and MSI. 

Performance

Testbed:

• Processor: Intel Pentium 4 3.0 GHz (200x15) with HT enabled, Socket 478
• i875P based mainboards:
• ABIT IC7-G (BIOS 11)
• ASUS P4C800 Deluxe Gold (BIOS 1004)
• Chaintech 9CJS "Zenith" (BIOS of 30.04.2003)
• Gigabyte 8KNXP Ultra (BIOS M8)
• Intel D875PBZ (BIOS P05)
• MSI 875P Neo-FIS2R (BIOS 1.3B2)
• Memory: 2x256 MB PC3200(DDR400) DDR SDRAM DIMM TwinMOS, CL 2
• Video card: NVIDIA GeForce FX 5800 Ultra
• Hard drive: IBM IC35L040AVER07-0, 7200 rpm

Software:

OS and drivers:

• Windows XP Professional SP1
• DirectX 9.0
• Intel Chipset Software Installation Utility 5.00.1012
• Intel Application Accelerator 3.0
• NVIDIA Detonator XP 42.82 (VSync=Off)

Test applications:

• VirtualDub 1.5.1 + DivX codec 5.05 Pro
• WinAce 2.2
• Cachemem 2.4MMX
• Gray Matter Studios & Nerve Software Return to Castle Wolfenstein v1.1
• Croteam/GodGames Serious Sam: The Second Encounter v1.07
• discreet 3ds max 5.1
• SPECviewperf 7.1

Here are brief characteristics of the today's contestants:

Every board in this round is a real piece of art, and you can choose what you need and like if you can afford it. Now let's have a look at their performance and then discuss it. 

Test results

The MPEG4 encoding speed is mostly dependent on the memory throughput, and in this test the difference is hardly noticeable between the models. The ASUS's and Gigabyte's boards look a tad better, and the MSI 875P Neo-FIS2R traces behind, but the maximum gap is only 3%.

The archiving rate much depends on another memory aspect - access delays, and the scores are very impressive. The breakaway of the ASUS P4C800 Deluxe is very great (21% from the last position). It's an abnormal difference for mainboards built on the same chipset (usually it makes 3-5%). Probably later we will look deeper into this mystery. As to the rest, we can see that the Gigabyte 8KNXP Ultra takes the second position with a 5% breakaway from the main group, the MSI's board outscores them by 4%.

The low-level memory tests prove that the ASUS is a cut above the others, and the MSI demonstrates an inferior read speed. Is this effect peculiar only to the synthetic tests and archivers? Let's have a look at the applications which will be very important for future owners of these boards.

In modern dynamic 3D games the ASUS keeps ahead of the main group of 3 boards by 5-7% in low resolutions and by 3-4% in high resolutions. The Gigabyte is a tad ahead, and the MSI is a little behind (2-3%). The gap in the games is not that impressive, but it doesn't disappear at all. 

The performance race

Let's leave aside the ASUS (it will be analyzed separately soon), and focus on the other boards. All the board have lifted up their performance with new BIOS versions. In particular, looking at the scores in the most critical test - archiving - we can see the following: 

• Gigabyte 8KNXP Ultra: 11:50 (F5, Top Performance Off) -> 11:48 (M8, Top Performance Off) -> 11:10 (M8, Top Performance On) = +6%
• Intel D875PBZ: 12:39 (B01) -> 11:42 (P05) = +8%
• MSI 875P Neo-FIS2R: 14:03 (1.2) -> 12:12 (1.3B2) = +15%

These changes look interesting because: 

• We managed to get the Intel's board started with the latest BIOS version at lower memory timings (2-3-2-5 against previous 2-3-3-5), though the resultant effect is not proportional to the boost we can witness. By the way, we couldn't bring CAS Latency (the most important parameter) to 2 on the Gigabyte and MSI boards.
• The MSI has no visible improvement; only the CPU-Z utility working under the Windows shows that the timings lowered - 2.5-2-2-5 instead of 2.5-4-4-5. Note that there are no traces of dynamic overclocking of the MSI board which is so popular these days (FSB clock speed (and the CPU's speed) goes up as the CPU load grows up, and then goes down to the default value when the CPU stands idle) - have a look at the rendering in 3ds max, the speed of which directly depends on the CPU's frequency and hardly on other factors.

 

• The Top Performance enabled in the BIOS in the Gigabyte's board lifts up the clock speed of the FSB and CPU by 1% (in defiance of what the engineers said), but the internal optimization of the memory controller  reveals one unpleasant fact.

Only in the test of the professional 3D accelerators, where both memory/processor bus and the memory/videocard (AGP) bus are loaded simultaneously, the scores of the Gigabyte 8KNXP Ultra slump in all subtests. I suppose the engineers made the AGP channel narrower in order to expand the dataway between the processor and memory (probably, by lowering the priority of requests from the video accelerator). Since the Top Performance setting doesn't affect the games, even in the heaviest modes (a large memory volume of modern 3D accelerators makes up for the AGP throughput), this setting of the Gigabyte board can be considered acceptable and even recommended.

Conclusion

So, the performance of these boards was raised up to the figure we fixed in the tests. It's only the ASUS P4C800 Deluxe which has a sound advantage over the others - it leaves the others behind by over 5%. The rest of the pack go on a par, with the Gigabyte 8KNXP Ultra being a little faster and the MSI 875P Neo-FIS2R a bit slower. 

Actually, it's impossible to choose a fast runner on the i875P chipset without overpaying for unnecessary accessories (except maybe the Intel's board). But if you know what FireWire is, if you are going to turn your disc array into RAID and make a transition to the Serial ATA, there is an excellent choice, - just forget about the price and look through the boards' functions in the middle of the review. All the boards look great regarding the number of overclocking settings and accessories (though it doesn't refer to the Intel D875PBZ). Especially I must note that the Chaintech 9CJS Zenith has the best accessory pack and the ABIT IC7-G and MSI 875P Neo-FIS2R have a perfect overclocking potential.