The problem appears to be caused by the way the Intel 975X mobo is reading and reacting to thermal data from the CPU and thus controlling the fan speed of the stock Intel CPU cooler, although these things get complex enough that I hesitate to pinpoint an exact cause.
I took a screenshot of the Intel monitoring utility showing fan speeds at about 3100 RPM while the CPU temperature was 88Cway too hot. We managed to banish thermal throttling by swapping in a Zalman CNPS9500 LED cooler with a manual fan speed controller, allowing us to finish our benchmarking.
After wrestling with the issue some more, I provided an update on things in our review of the Extreme Edition 955. By then, I was leaning toward motherboard warp as a possible cause of the problems:
Chris speculated that the CPU cooler might be tensioned too tightly, causing the Intel D975XBX motherboard to warp and thus preventing the cooler from making clean, even contact across the surface of the CPU. This theory would seem to explain why we saw relatively better (though not entirely satisfactory) results when using the cooler’s included TIM pad on initial installation of the CPU rather than thermal grease.
This seemed like a solid explanation, especially since the 955XE consumes less power under load than its predecessor.
After the publication of our 955XE review, Intel sent out a third, “production level” motherboard/CPU/cooler combo for us to try. (Major kudos to Intel for following up there, by the way.) The changes of note in the production version, they said, are modified VRMs on the motherboard and a different cooler that’s tuned for thermal performance rather than optimal acoustic performance. This cooler responds to motherboard input with higher RPMs than the other one. It seems that review kits somehow went out with the wrong coolers, tuned primarily for acoustics rather than thermals. Retail boxes with Extreme Edition 955 processors will ship with this thermal-tuned cooler instead.
With the possibility of motherboard flex in the back of my mind, the first thing I noticed about the new cooler is that it has a smaller copper surface area to come into contact with the CPU, possibly mitigating problems with fit.
Securing the cooler’s four retention pegs seems to require less downforce, as well.
I have to admit, though, that mobo warping problems were swept out of my mind when I fired this thing up and heard it whir to life. The cooler eventually settled into a reasonably quiet speed once the system had POSTed and finished booting into Windows. But when I subjected the 955XE to our torture test of four simultaneous Prime95 sessions, the fan quickly ramped back up into high speed again. The final result: the CPU runs stable with no thermal throttling at about 71-72C with the fan cranking at approximately 4200 RPM.
Under these conditions, this 955XE setup is significantly louder than any other CPU/cooler combo we’ve tested in the past few years, registering on our decibel meter at 62.6 dB from about one foot away. By contrast, our Athlon 64 X2 4800+ test system weighs in at about 46 dB. That’s a big difference in terms of decibels, and our ears confirm it. Yow. This aural performance is somewhat puzzling because the combination of the Extreme Edition 840, Intel D955XBK motherboard, and Intel cooler was both adequate and relatively quiet.
If you plan on building a system with a 955XE processor, I’d say an aftermarket cooler is almost a requirement. The Zalman was able to achieve lower temperatures at its slowest fan speed than this production Intel one does at 4200 RPM. The retail Intel cooler at least does its job properly, but it’s way too noisy.