As a result, I've updated the review with new 3DMark scores for the X1950 CrossFire rig. With the new scores, the Radeon X1950 CrossFire system opens up a distinct lead on the GeForce 7900 GTX SLI system in 3DMark.
These new scores don't alter my overall conclusions about the Radeon X1950 XTX, but some additional graphics testing work I've done over the past couple of days has underscored how nice it is to have some of the Radeon X1000 series' features in an ultra-high-end GPU setup like this X1950 CrossFire rig. With that amount of raw pixel-pushing power, you'd definitely want to use 16-bit floating-point HDR texture formats in combination with antialiasing. You'd want to have the highest quality anisotropic filtering algorithm available. (ATI's is superior by default to Nvidia's and only gets better when you turn on their high-quality angle-independent aniso feature, which the G71 inexplicably lacks.) You'd also want really good antialiasing capabilities, and ATI has led Nvidia on that front since the introduction of the Radeon 9700. In short, ATI has a good thing going here that current GeForces can't match in terms of peak image quality. If I were seriously looking at plunking down the kind of money involved in a high-end dual GPU setup, I'd consider that fact at length before making any purchases.
As for the D975XBX motherboard issue, I'm not sure what the story is there. The problem only seems to affect two tests in 3DMark06, where the slave card appears to be dropping out of the mix. I tried about every possible permutation of BIOS revisions, DIMMs, Radeon cards, power supplies, video drivers, and the like that one could imagine, to no avail. I was able to get the CrossFire X1950 working on the Asus P5W DH and on an older revision of the Intel D975XBX. (Unfortunately, the older D975XBX won't support a Core 2 processor, so I couldn’t get comparable scores out of it.) That makes me think perhaps my copy of the new-rev D975XBX might have been the culprit.
Whatever the cause, I'm currently re-running some of our other tests to confirm that the D975XBX didn't slow the X1950 CrossFire config's performance anywhere else. So far, I have no indication of that being the case.