URGENT HEADS UP
I, myself, just found the X-Bit Labs review, and drilled down to the relevant material on pages 5, 6 and 7.
We all know X-bit Labs ? they often produce good benchmarks of products with reliable data and conclusions. However, now I have doubts.
The table on page six only provides load temperature values. We would like to have the idle values, from which we could compute thermal resistance given the TDP posted at Intel for the 521 processor. I'm still searching for reseller descriptions of the 521, and should probably go directly to Intel. But the 520 processor is a Prescott core running at 2.8 Ghz; the 531 processor is a Prescott core running at 3.0 Ghz.
The reviewer says he over-clocked the test-bed to between 4.0 and 4.2 Ghz, but this would represent a 50% over-clocking above the stock values. I have not heard of such record-setting increases in stable clock speeds with either air or water-cooling systems. I believe I recalled reading of cases where 40% over-clocks were achieved with Vapo-Chill -- keeping the core cooled to near freezing. Please correct me if my seat-of-the-pants recollections are incorrect.
Now the 531 processor has a stock front-side-bus speed of 800 Mhz (hence the 521 would also have the same FSB), although the test-bed motherboard is capable of 1066 Mhz. The reviewer cites using Corsair TwinX1024-4300C3 modules, which would be equivalent to DDR2-533's. If it were even possible to get an over-clock of 50%, he would have to choose a divider different from 1:1 unless he were running these modules at 600 Mhz and beyond their rated speed.
For further comparison, my XP120 ThermalRight cooler has a thermal resistance of 0.167, installed on a Northwood 3.0C processor which has a TDP value of around 80 watts. This would give an idle-to-load temperature spread of 13C, and I have confirmed that independently -- indeed, the difference between my load and idle temperatures is 13C degrees -- well, more like 12.7C. Since the Zalman 7700 cooler has a thermal resistance of 0.19, I would expect a similar drop in load temperatures for my XP120 and Northwood processor, which seems ridiculous: there would be no variation of load from idle temperature at all.
While I have not done the careful math, and I'm comparing my Northwood with the Prescott core in the review, I am reserved about my statements in the paragraph above in regard to that comparison. But the TDP of the test-bed processor cannot be greater than 120 watts. So you would expect that a 13C drop between the CNPS-7700 and the CNPS-9500 would translate into a 13C drop for my Northwood -- at least that.
I believe we have stumbled on a bogus review designed to hype the Zalman CNPS-9500 cooler.
I encourage additional comments here, and I would agree with someone who posted previously to say that we want to look at several reviews before we agree that this cooler is really as great as either Zalman or X-Bit Labs says it is.
What frightens me most here is that X-Bit Labs had previously commanded more trust from me. I thought I had been impressed earlier at their scientific approach in these matters. Perhaps it is a matter of the way they obtain their reviews, and this time, a bad one slipped past.
Wake up, X-Bit Labs!
Facebook
Twitter