nomenclature sports
jd
case = IHS in CPU parlance (NOT internal case air temps - as JoeC often refers to these air temps)

sink C/Ws are described as 'case to air' or 'sink (DUT bp) to air'; the values are different because the specific test procedure includes or excludes the TIM joint between the IHS and bp

if a semblance of comparability with 'industry methods' is desired, I suggest a fine slot on the face of the die for a 40ga type T thermocouple
-> this IS the 'case temp' as measured by the big guy, the little player grooves the sink bp so has only 'sink (bp) to air'

Testing is relative - If I use only one pump at one flow/pressure rate I will get one set of numbers that do not reflect the real or total picture as they all have diff pressure drops High vs Low. so you test a range that defines performance.

Now if Intel, AMD both use an IHS which introduces a whole new layer of resistance, ignoring that assumes that all thermal solutions will test the same with an IHS. Can you validate that? I would be most interested

An IHS is able to spread the heat at least somwhat. In the ""Thermosyphon Independent Tests"" OC posted I am guessing at a bias in the the comparison to the XP90C as it performed better on large die and IHS in system. This why one stop single die testing without considering the total picture is an inaccurate representation of the facts

Testing die only is accurate but missleading as you only have one point. of reference. OC's Multiple point tests tell you that such and such a product is good in one situation, but maybe not as good in another. I do not think OC's data presentation is well organized and can see where improvements could be made but all in all better than most.

Interesting:
http://www.overclockers.com/articles1259/

a good description of the difference between his benches

Appalling.

what ?
the difference
or Joe's description ?

Agreed. Kinda skewing data to prove a point there.
Want to prove anything, you have to actual test and get data. Only estimates.

2)He's overlapping premuim heatsinks vs. underperforming WCing kits that are also 2-3 years old like a innovatek kit with eheim 1046. Not really a fair comparision. If he compared a Dtek kit for $170 vs $50 heatpipe thats a 3.5x cost factor, not 5x. I even checked out a custom kit for $120-$150 that would outperform heatpipes at 4800 rpms. So they are not that expensive.

Yet his conclusion says it all. "All my desktops are water cooled, first and foremost for low noise. Newer heatpipes can effectively compete on this basis and deliver extremely good cooling performance, assuming excellent case airflow. Water will always trump air for the load it can handle."

I geuss the APEX kit did a lot better than he let off on his review.
And I thought I was being paranoid and his test results were way off and he had some motive making the kit seem so second rate.

You, I, and many others, spend the last 4 years championing coherent and correct testing through minimising variables and now here's JoeC who many people seem to respect, setting internet testing back by 3 years in upholding abstract interpolation of datum as a means to extract "conclusive" recommendations on comparitive performance.

If he wanted to prove a point, pick a top, middle, and lower performing items of the water and air-cooling world and retest those on a singular test bed rather than mish-mash it all together into some abstract "corrected" ranking.

One thing for a forum member to extrapolate datum from different testbeds when no such directly comparitive information otherwise exists. Quite another for someone who purports to independently assess cooling performance of products to do the same when the tools to make a direct comparison are sitting in front of him.

May as well have not written anything.

Got to admit that one has my eye twitching.


He ought to pull the article and take a vacation I think he needs to replace that monitor tan with a real one

I took it for what it was worth, but what really spooked me is the numbers:

The best kit tested out with a C/W of "0.05". No error margins are stated, and with that kind of figure, it leaves a huge question about the results. If the error margin is +/- 0.01, then the order of the top 4 could be re-arranged any which way.