Finally!!! An Accurate Heatsink Comparison!

[Laz]

Their meathodology may be just as flawed as other MB sensor reviews, I said "may" but that may not be true. I don't believe one should be trying to draw cause and effect conclusons from corrolations, it could easily be a different variable that affects the results observed, such as the size of heatpad used (being larger in surface area, compared to a cpu core), or the pressure applied to the heatsink, or even conditions inside the case where the heatsink is mounted. These are just a few variables and I'm sure there are others. Most reviews are simply trying to give the reader a ranking of the efficiency of each heatsink as opposed to an absolute effectiveness. I havn't really had the time to evaluate their procedure as I'm reading it from work and that is why I stated "may" be flawed, but I believe most reviews that do a simple comparison are able to convey the correct ranking to the reader.
Having said that, it would obviously be more accurate to drill the sink and measure the temp using a more accurate reading, while the sink is on the core, as I believe was done in the link you provided. This would be more of an accurate reading in absolute terms, but I don't believe the relative ranking of the heatsinks would change, and if they do, then my reasoning is flawed and I would agree that the MB temp reading is useless and should never be used for comparissions of heatsinks. As you can probably tell I'm not that familliar with the accuracy of the MB monitor, but I'm assuming it gives the correct "relative" reading, and it's my basis for assuming it can be used for reviewing "relative" ranking of heatsinks, albeit not accurate readings.
I still don't see the purpose of using the large heatpad to heat the sink, or the neccessity of testing the sink outside it's intended use. I would prefer to see the sinks tested in a computer, running the exact same programs, using the exact same conditions (temp, duration, etc.), and obviousy if the reviewer has the equipment, it would be nice to drill the sink and use a more accurate measurement. If I misunderstood the process, I'm sure you will inform me and I will try to reread it when I get off work and I have more time to try and understand their reasoning behind their meathods.

 

[Mikewarrior2]

The overclockers.com was not intended as a heatsink test, as i've been told, but rather an alternative, more accurate method to reading cpu temp for tests.

The fact remains that socket A readings are off. Yes, the order is probably correct, but if you are a beginner user, and see a chrome-orb(sells for under $15) coming within 2C of an alpha pal6035(sells for $30), then that new user is probably gonna buy the chrome orb.

Chances are that won't effect any people on the bbs, because we all know that the orb series doesn't really perform all that well. But for the beginning computer user, who reads solely hte front-page of the website, and see's such a miniscule difference between alpha and c-orb, he'll probably get the c-orb. And without at least amending the article saying that temps are accurate, this is a mistake.

There is a reason the c-orb is only rated to 1ghz. At 1ghz(1.8V for overclocked Chip), cpu CORE temp would be around 60C. Unfortunately, the mb sensor reading is low-balling this number by anywhere from 10-12C. So you get the issue of both accuracy, and whether or not this "relative" test is a valuable test.

I feel it only serves as confusion. YOu have 6 or 7 Heatsinks all bunched in within 3C of each other in the Anandtech test. When in actuality, between those 7 heatsinks, chances are temps are in an 18C range(from highest to lowest). Relatively, they are in order of performance, probably. There has to be some guesswork for those heatsinks that tied.

So to the unknowing user, all 6 of these heatsinks all perform generally the same, but that really isn't the case if you're accurately measuring core temp.


Mike