Anandtech Socket A heatsink review

[Laughingman]

Better Heat Sink Test Comparison

Similar test method to Intel spec. Compares Motherboard monitor to calbrated thermocouple should answer a few questions.

Regards

 

[Mikewarrior2]

Thanks to everyone in my sig, and Laughingman, JohnCar, Nevin, etc.

*Bumping the thread for those who haven't seen this new test.


Mike

 

[Laughingman]

Calibrate your Mother Board monitorFrom Burning Issues. A useful way to calibrate your Hardware monitor systems for relativley Accurate Temperature readings. Big thanks to our UK brothers.

Regards

 

[HellRaven]

I have read through the entire thread again and I must say I support you guys 100% This is an issue across the entire internet wherever a heatsink is reviewed, but like Mike said - from Anandtech I expect better. Tillman came into this with good intentions and that is fine, but I think it has been proved that his methods were flawed. People screw up, it happens (even on anandtech ).

Tillman, Anand, anyone else - Now it is time to fix the problem. Don't leave that article up there where those less informed can continue to read it and draw the wrong conclusion.

 

[Dexion]

I completely agree with Mikewarrior, Laughingman and Johncar. Kudos for your distinct effort here! Definately this thread is a MUST read for those who follow cooling to a T(Pun intended). They have officially proposed a standard(even Intel and AMD use) for all heatsink testing methods for all hardware reviewers. It up to the harware review sites to follow or not, and those that do are the ones that are worth reading.

 

[VladTrishkin]

Its too bad all the hardware sites will be doing it wrong all over again.

 

[Tillmann]

Hello,

I haven't been checking the forums in the last days, but that does NOT mean that I've lost interest. I take your criticism very serious, and I see that you guys are obviously waiting for some kind of statement from me.

I will be looking at different testing methodologies in the next days - I'll also contact Joe from overclockers.com about this tomorrow. Maybe I will use test equipment similar to the equipment presented in the overclockers.com article, maybe not. I'll decide that when I have more info from Joe. There are a few things that intrigue me about the overclockers.com test - e.g. why does Global WIN FOP32 outperform the hedgehog on an Asus board, but not on an Abit board (normally heatsink performance shouldn't depend on the motherboard). But I would not like to have a public discussion about that; I'm not interested in confrontation, but cooperation. The goal is to work out a heatsink testing methodology everybody is happy with.

Concerning the recent roundup: Anand and I have agreed that there will be no changes to the roundup, unless the new test shows significant differences in the performance/standings of the units. If your claims about the inaccuracy of the current testing methodology are true, this will be the case, otherwise not. But even in the case that there should be no significant differences, I will publish detailed information on the testing methodology and test results (possibly on heatsink-guide.com and not AnandTech).

bye,
Tillmann

 

[Lightingguy]

Hi!

You certainly opened Pandora's Box on this subject. Seems like the situation is somewhat parallel to the quantum mechanics theorem that states by attempting to measure something, you unavoidably affect it...and that the more accurate your attempted measurement, the greater the margin of error you introduce.

Having said that, I certainly applaud your efforts and those of the many others who have added their opinions, crash courses on thermodynamics, thermocouples, miscellaneous devices, procedures, opinions, etc.

I respectfully submit that, short of the quantifyable heat dissipation potential of the tested devices, in-situ testing of HSF is a useful but innacurate benchmark due to the (noted) variations present in hardware monitoring devices on MBs. In addition, the method of mounting, thermal pad/paste/grease application, and other factors may, in fact, have a greater bearing on just how these units operate within systems. That's why some people have great results with Super Coolers, and some curse their Hedgehogs.

Sorry if I run on. IMHO, this is the place for hashing out not only products, but testing methodology as well. Keep it up!

 

[johncar]

Lightingguy wrote...quote...
I respectfully submit that, short of the quantifyable heat dissipation potential of the tested devices, in-situ testing of HSF is a useful but innacurate benchmark due to the (noted) variations present in hardware monitoring devices on MBs. In addition, the method of mounting, thermal pad/paste/grease application, and other factors may, in fact, have a greater bearing on just how these units operate within systems. That's why some people have great results with Super Coolers, and some curse their Hedgehogs. end quote....

We also respectfullly and heartily agree. And let us demonstrate the validity of this point of view by considering similar attempts to test cpu to hs interface materials with in-situ testing. We are all aware of "review sites" contesting the effectiveness of silver grease and esp the latest product Arctic Silver over zinc oxide grease, using in-situ testing with above mentioned inaccurate mobo hardware monitoring devices and procedures, and absolutely no idea of cpu power dissipation. This is the height of foolishness and "self interest of review sites", when we know from credible published material properties that silver greases are typically ~11.5X more thermally conductive than the Radio Shack Zn0 grease. Point...why in heavens name does anyone need to conduct these inaccurate in-situ tests. One can more accurately "calculate" the reduction in die temp for any estimated or chip mfr's max spec power dissipation. Review sites would be better advised to publish appropriate tables based on these data.

Problem with "heatsinks" is that mfrs' tests of thermal res, C/W, are not necessarily available, nor necessarily standardized, leaving the door ajar just enough for third party intervention, however inaccurate. Maybe someone should collect and publish whatever mfrs' test data and procedures may be available.
John C.

PS:-A few years ago, Aavid, a well respected long term player in the electronic thermal solutions arena, designed a very professional experiment/apparatus to test the "Thermal Performance of Interface Materials in Electronic Packaging Applications"...happens to be the title of the paper authored by Messrs Early, Lee, and Pellilo, selected from following page.
http://www.aavidthermalloy.com/atp/techpaper/techpaper.html

Unlike the many "young Tom Edisons" who rely on one improperly located thermistor by mobo mfrs, the Aavid engineers placed interface materials between extemely highly polished surfaces of 1 sq in aluminum rods, completely insulated, metered heat flux, and accurate temp measurement techniques to measure the temp drop across those polished surfaces which had as closely as possible no temp differences along either surface... ie isothermal surface temps. They also controlled and measured contact pressure. Worth a read to learn what it takes to design an accurate experiment.

 

[Mikewarrior2]

Tillmann,

Multiple people with far more experience, training, and knowledge in
thermodynamics have already explained, shown, and demonstrated beyond a
reasonable doubt why your testing method was invalid. Your determination of
whether it was valid and any attempt to defend it is irrelevant.

Your continued belief that your heatsink review is "accurate" is pretty amusing, but nonetheless wrong.


Mike

 

[Mikewarrior2]

Thermite88,

Since I did not reply before to your point, I will do so now. The amount of heat loss at the backside of a cpu, per the overclcokers.com tests that isolated a thermocouple directly behind the core and sandwiched it with closed cell foam, the temp loss on the backside from directly behind core to core edge(where kt7 thermistor's measure temp), is roughly ~25%. Ignoring the isolation technique that overclockers.com uses(i'm not sure if this is the proper method or not), i would say that, from the position that the kt7's measure temp, the heat loss versus primary pathway is between 25-30%.

Assuming that the backside core temp is within 1%, you're still very significantly off in terms of heat loss at where the kt7 measures temperatures. Of course, I do no tknow how close the direct backside temp is in comparison to actual core temp.


Mike

 

[thermite88]

Mikr,

Read johncar's comment on Nov/24/2000 5:39 PM. That should help you get a better feel about what it takes to get accurate measurements. It is easy to measure temperature, but it is very difficult to derive heat flux from uncontrolled temperature measurements. I don't really agree with your reasoning.

 

[Mikewarrior2]

Reading JohnCar's post again, I don't understand why you don't understand my reasoning?

It would seem rather obvious that an uninsulated thermistor reading(ie in the Anandtech review) would have a greater than 10% drop from the temp read at the primary heat pathway? Judging at the overclockers.com article, where backside temp between back of core versus back of core-side, the temp difference is roughly 25%(40C for core-side, 50C for core directly beneath).

Isn't that enough to put doubt into the validty of the uninsulated thermistor's readings? I would have to say that the difference between primary heat pathway versus secondary heat pathway(in the way socket A temps are measured) are at least 10%, if not greater.

Also, you have me at a definite disadvantage in this argument. I can't argue about heat flux one way or the other, as this is something I don't really yet know a lot about. I definately understand the need for isolating/insulating the thermistor/thermocouple in order to minimize secondary heat flux.

Seeing as I don't personally have the resources to conduct backside heat flux tests, I can't really comment on your first post in this thread. I will have to leave your question still unanswered until I either learn more about this or see an experiment like this done.

Sorry I could not fix the blanket statement over your post. I dont' have the knowledge to answer it. I would greatly appreciate any additional information. I am just here to try ot learn more about it.


Mike
P.S. On the issue of the Tomshardware review, the Taisol(.48C/W) and PAL6035(.38C/W) were seperated by 1C, so we still see the "compression" of temps.