CPU: Intel i7 i950 and i970
Motherboard Bios Versions:
- Asus Rampage III Extreme bios version 0704
- Gigabyte UD9 F2 bios
- MSI Big Bang XPower Bios version 1.2
Video Card: eVGA GTX 285 1Gb videocard ( WHQL 258.96 )
Memory: 6Gb Corsair Dominator 1600CL8 or 6Gb Corsair GTX 2 2250CL8 ram
HDD: Western Digital 1Tb Caviar hard drive
PSU: Enermax Galaxy 1kw DXX PSU

Cooling Solutions Used on the CPU:
Air : Thermalright VenomousX with dual 1600rpm fan
Water : EK HF Supreme waterblock, Thermochill  PA120.3
Phase : Vapochill modded by Jort Casing : Lian Li PCA-77

Synthetic Benchmarks

Time to put the boards against one another. As previous roundups  have shown, there isn't much to expect when comparing similar chipsets clock for clock. Though I was looking forward to see how the UD9 performed compared to the two other non NF200 equipped boards. We expect a slightly lesser efficiency. Starting with the normal synthetic benchmarks. Later we will test max bclock, ram clocks and a short recap on B2B delay. The latter for the extra stability it offers when running large amounts of ram.

As usual we start off with SuperPi and Wprime. Especially superpi 32M gives a good indication how efficient a board can be. The 1M test is more about CPU speed and might be a bit too short in run time, to decide who is or who is less efficient. Wprime is also more CPU bound, so we expect little to no differences. Let's see what happens:

The REX III is staying nicely in front, we were surprised to see it even take a small gain in these short tests. For all the benchmarks, only the required onboard hardware features were enabled (LAN ports, sound, etc were disabled) It's nice to see the progress against the first generation REX II board, generation two is better dialed in and optimised. Wprime as expected is pretty close. Note that all the results are the best run out of 3.

Superpi 32M however is another story. Time after time the UD9 proved to be the quickest, though the MSI XPower board is right on its heels. The REX III beats it's predecessor, though the fast run we saw at Superpi 1M isn't reflected in the 32M run. Wprime 1024 seems ASUS territory, the two Extreme boards just staying ahead of the two rival boards.

Time to throw some 3D into the action, Futuremark's 3D01 remains still a great test for the subsytem. 3DMark06 as it gives a good reference towards GPU and CPU score. Thirdly I added the results of the overall score of the PCMark05 runs. The latter giving a very good indication of the overall system performance.

Let's analyse the results now. 3D01 is fastest on the MSI board. Maybe the lack of added gimmicks on the board makes it the faster board here. The ASUS duo is very close and the UD9 limping a little bit behind. We've seen a similar performance loss with the eVGA Classified 4 Way board. Maybe the added NF200 put some strain on the 3D hardware. Making it perform a tiny bit less. The Rampage III Extreme takes the 3DMark06 crown, just beating MSI's XPower board. The UD9 takes a split 3rd place with the Rampage II Extreme. The differences aren't that big, but there was a clear trend that in 3D the UD9 slightly trailed the pack. The same counts for PCMark05 where it's the REX III, XPower 2nd and settling for 3rd spot, the UD9. Finally, last but not least : the REX II board. As predicted, really nothing earth shattering, a measly 100-200Marks difference. For benchers maybe valuable information, for daily users or abusers not much you would notice in real world applications.

The Western Digital 1Tb Hard drive was hooked up to the ICH10R controller. The full HD Test suite was run to see if there were any discrepancies in the file transfer department. Really a close call again, as the boards are neck to neck. A clear winner is hard to determine.

Cinebench up next. Testing the I950's single and Hyperthreaded enabled (8) cores.

Performance is similar across the tested boards. Not much to see here, is there ? Let's move onto Everest...

This is not my favourite benchmark as sometimes results can be way off. But since most use it as a guideline we stick to it; Though I interpret the results as so so, sometimes boards giving great Everest Results are under performing in eg Superpi or other bandwidth related tests. A tiny bios setting can influence the output big time. So we setup all boards alike.

Only the Copy result of the Xpower board was lower than the other 3 boards. For the rest, same story. The Asus boards pulling ahead, but as I said before : this result is contradicting the Superpi 32M score. I have more belief in the validity of the Superpi 32m score. Bios optimisations play a big part.

Next page we continue with some game benchmarks, which might be more interesting than the synthetic stuff...