Finally!!! An Accurate Heatsink Comparison!
- Thread starter Mikewarrior2
- Start date
heheh, now you dont have to explain how motherboards monitors are off, just put up a url ;-)
Nothing like seeing my hedgehog haul ass for a second time ;-) thanx for the link ;-)
Nothing like seeing my hedgehog haul ass for a second time ;-) thanx for the link ;-)
Correct me if I am mistaken, but according to the accurate thermocouple results of the test, the Millennium HSF came out on top.
Hedgehog ABIT KT7 Thermocouple 22.1C 35.4C
Millenium ABIT KT7 Thermocouple 19.5C 33.0C
Hedgehog ABIT KT7 Thermocouple 22.1C 35.4C
Millenium ABIT KT7 Thermocouple 19.5C 33.0C
Hmm, the results didn't differ greaty from AT's review. The numbers were a tad off, but the Hedgehog came out on top in each, followed by FOP-32, then Alpha Pal. I don't see what the fuss is about. And by-the-way, I used Motherboard Monitor 4.17 to find my cpu temp in another thread and you said, it CAN'T be running at the temp it gave me. Aside from examples where the heatsink didn't fit, MM 4.17 gave within 2 degrees Celsius of Thermocouple according to there review.
Yeah, but obviously the guy who did the test was buds with the millenium guys, so i ignore those results ;-)
Jaydee,
There are significant differences between thermocouple and mb readings.
For example,
Asus Probe, FOP32... OnMB sensor, 44.5... Thermocouple, 35.5 A difference of 9C. I don't get where you see only a 2C difference.
Abit Thermistor, FOP32... OnMB sensor, 41C. Thermocouple, 36C. a difference of 5C.
Not to mention, another 9C difference with the Hedgehog + A7V measurements.
So before you claim that I was wrong, or that hte Anandtech review was right, open your eyes first. Comparisons with on-mb sensors cannot be done, plain and simple. And the sensors are off, period. The problem is, you can't predict which way they are off, either high or low off.
Mike
Oh, And here's another one. Millenium + KT7, 16C difference.
Oh, and the unpredictability of which way the thermistor reading is wrong is yet another reason to not compare heatsinks with an on-mb sensor.
There are significant differences between thermocouple and mb readings.
For example,
Asus Probe, FOP32... OnMB sensor, 44.5... Thermocouple, 35.5 A difference of 9C. I don't get where you see only a 2C difference.
Abit Thermistor, FOP32... OnMB sensor, 41C. Thermocouple, 36C. a difference of 5C.
Not to mention, another 9C difference with the Hedgehog + A7V measurements.
So before you claim that I was wrong, or that hte Anandtech review was right, open your eyes first. Comparisons with on-mb sensors cannot be done, plain and simple. And the sensors are off, period. The problem is, you can't predict which way they are off, either high or low off.
Mike
Oh, And here's another one. Millenium + KT7, 16C difference.
Oh, and the unpredictability of which way the thermistor reading is wrong is yet another reason to not compare heatsinks with an on-mb sensor.
Oh, and Jaydee, you claimed a 21C Cpu temp, which is nowhere near the overclockers.com temps.
Mike
Mike
Ok, I was looking at Heatsink Motherboard Results, on page 3. I shoulda said I was reffering to idle temps only (my bad, but it was because I was asking for idle temps in the thread I started). 4 out of the 6 idle temps in that category that the heatsink actually fit, were within 2 C, and ALL of the Abit boards that fit (which is what I have).Oh, and
<< Jaydee, you claimed a 21C Cpu temp, which is nowhere near the overclockers.com temps. >>
Alpha PAL6035 ABIT KT7 MBM 417 21C
Alpha PAL6035 ABIT KT7 Thermocouple 22.9C
GlobalWin FOP32 ABIT KT7 MBM 417
GlobalWin FOP32 ABIT KT7 Thermocouple 22.2C
Hedgehog ABIT KT7 MBM 417 21C 3
Hedgehog ABIT KT7 Thermocouple 22.1C
And again, they reached the same conclusions as AT, at different temps, but the same h/s came in at the same order.
If you had such a problem with the article why didn't you post it in the AT Articles topic?
<< Jaydee, you claimed a 21C Cpu temp, which is nowhere near the overclockers.com temps. >>
Alpha PAL6035 ABIT KT7 MBM 417 21C
Alpha PAL6035 ABIT KT7 Thermocouple 22.9C
GlobalWin FOP32 ABIT KT7 MBM 417
GlobalWin FOP32 ABIT KT7 Thermocouple 22.2C
Hedgehog ABIT KT7 MBM 417 21C 3
Hedgehog ABIT KT7 Thermocouple 22.1C
And again, they reached the same conclusions as AT, at different temps, but the same h/s came in at the same order.
If you had such a problem with the article why didn't you post it in the AT Articles topic?
Hmm, that's pretty cool. I wish I could afford that kind of equipment to test my system 
.
Pretty surprising how better the FOP32 turned out to be compared to the PAL6035, and how close it came to the hedgehog. All the reviews I've ever seen have either put the FOP32 in a tie with the PAL or a little worse, but never better.
.Pretty surprising how better the FOP32 turned out to be compared to the PAL6035, and how close it came to the hedgehog. All the reviews I've ever seen have either put the FOP32 in a tie with the PAL or a little worse, but never better.
A couple of comments about the testing:
C/W is C/W. In our tests we tried to be as consistent as possible. What should have been brought up in the testing review was the relationship between our results and what manufacturers data was available.
The Alpha Pal manufacturers data .38C/W Our data .37C/W - .4C/W
The FOP 32 manufacturers data .3015 C/W Our Data .235C/W - .266 C/W
The Agilent CA manufacturers data .48C/W Our bench test .34 C/W subsequent GiGabyte board test .45C/W
CPU 51C - Ambient 24C
Understand that this was an article about heat sink Testing, not heat sinks. Our indicated test error was generally in the 10% range with one or two errant points.
Some variables such as MM reported Core Voltages were suspect. Additionally type of grease (Arctic silver) Clamping pressure, Manufacters understatement could have all contributed some error here.
Our intent here is to develop a reliable method (and a practical one) for testing heat sinks. This is an ongoing development Incorporating refinements as we proceed. Tillman or anybody else that would like to contribute to our effort is most welcome.
I think the criticism of tillman was a little overdone although I did jump in on it, most was unjustified.
Regards
Andrew Lemont
Millennium Thermal
with thanks to Joe and Stephen
C/W is C/W. In our tests we tried to be as consistent as possible. What should have been brought up in the testing review was the relationship between our results and what manufacturers data was available.
The Alpha Pal manufacturers data .38C/W Our data .37C/W - .4C/W
The FOP 32 manufacturers data .3015 C/W Our Data .235C/W - .266 C/W
The Agilent CA manufacturers data .48C/W Our bench test .34 C/W subsequent GiGabyte board test .45C/W
CPU 51C - Ambient 24C
Understand that this was an article about heat sink Testing, not heat sinks. Our indicated test error was generally in the 10% range with one or two errant points.
Some variables such as MM reported Core Voltages were suspect. Additionally type of grease (Arctic silver) Clamping pressure, Manufacters understatement could have all contributed some error here.
Our intent here is to develop a reliable method (and a practical one) for testing heat sinks. This is an ongoing development Incorporating refinements as we proceed. Tillman or anybody else that would like to contribute to our effort is most welcome.
I think the criticism of tillman was a little overdone although I did jump in on it, most was unjustified.
Regards
Andrew Lemont
Millennium Thermal
with thanks to Joe and Stephen
Huh,.......................seems like even if the temps are not precisely correct w/mb cpu sensor, as long as the hsf test is performed identically w/each product installed on the SAME mb one after the other, it is still a good baseline comparison between hsf's.
Kind of like having a torque wrench that may not be calibrated precisely correct, as long as you use the same wrench, each bolt will still be torqued to the same reading.........(Just an analogy?)
Know what I mean.............................Vern?
Kind of like having a torque wrench that may not be calibrated precisely correct, as long as you use the same wrench, each bolt will still be torqued to the same reading.........(Just an analogy?)
Know what I mean.............................Vern?
Vern
Not true, check a little closer. Things are not what they seem at first glance. Also check the other threads on subject.
Regards
A
Not true, check a little closer. Things are not what they seem at first glance. Also check the other threads on subject.
Regards
A
JayDee, I have already made a thread in GH about this, and have brought it to the attention of Tillman, Anandtech's article writer.
Mike
Mike
clumsum,
Because the temps are unpredicably off, and the results are almost always "compressed" readings. One person may have a thermistor that reads 5C too high, another might have one that reads 16C too low. Too unpredicatable to go on "same mb, so we can compare temps argument" and too unpredicatable for tests even on the same mb. Relative "tests" don't work either.
Jaydee,
KT7, FOP32, 5C difference. Also, impossible to predict the difference. He may have gotten a 2C difference, doesn't mean you will.
Mike
Because the temps are unpredicably off, and the results are almost always "compressed" readings. One person may have a thermistor that reads 5C too high, another might have one that reads 16C too low. Too unpredicatable to go on "same mb, so we can compare temps argument" and too unpredicatable for tests even on the same mb. Relative "tests" don't work either.
Jaydee,
KT7, FOP32, 5C difference. Also, impossible to predict the difference. He may have gotten a 2C difference, doesn't mean you will.
Mike
So,................I think you are saying that the cpu temp sensor on the SAME board will not consistantly report the temp?
I'm not talking about the same brand and model of mb................I'm saying the ONE same mb!
Mount different hsf combos on one same mb, one after the other and run exactly the same tests....................for a base line.
KISS!
I'm not talking about the same brand and model of mb................I'm saying the ONE same mb!
Mount different hsf combos on one same mb, one after the other and run exactly the same tests....................for a base line.
KISS!
No, even the same board will not accurately portray cpu temp.
Besides the points brought up in the anandtech gh forum(like the alpha to c-orb readings only a 2c difference, but an actual difference of around 12C), and the overclockers.com article, the thermistor can read either high or low, even on the same mb.
For one heatsink, the cpu may read 10C too low, the other, about 4C too high. The result, the difference between teh two going from 16C, to 2C.(for example only). ANd the unpredictable nature, and general inaccuracy of the thermistor solution prevents even "relative" tests.
Mike
Besides the points brought up in the anandtech gh forum(like the alpha to c-orb readings only a 2c difference, but an actual difference of around 12C), and the overclockers.com article, the thermistor can read either high or low, even on the same mb.
For one heatsink, the cpu may read 10C too low, the other, about 4C too high. The result, the difference between teh two going from 16C, to 2C.(for example only). ANd the unpredictable nature, and general inaccuracy of the thermistor solution prevents even "relative" tests.
Mike
What I'm saying is the temp does not have to be accurate, as long as its the same sensor thats reading the temp each time..................?
The temps of the hs will vary by factors such as alloy composition, tightness of fit, speed and cfm of fan etc,....etc., but the temperature measured w/the mb sensor is underneath the cpu, so how effecient the hsf is working (for whatever the reason) will still be reflected by the cpu sensor......................?
The temps of the hs will vary by factors such as alloy composition, tightness of fit, speed and cfm of fan etc,....etc., but the temperature measured w/the mb sensor is underneath the cpu, so how effecient the hsf is working (for whatever the reason) will still be reflected by the cpu sensor......................?
No, that doens't work... For each heatsink, the temp may be off either high or low, even on the same MB/Sensor setup.
For example, look at the This LInk In GH.
In the Anandtech Review, the difference between the pal6035 Alpha and a c-orb was a mere 2C, when the actual core temp difference is around 12C or higher.
Much more detailed information from experts like JohnCar, LaughingMan, Shoarthing, etc. in that thread, but the bottom line boils down to thermistor inaccuracy, and the impossibility of running a comparison on a socket A platform.
Of course, if you, like Tillman, believe that a 550% comparison error is perfectly fine, then I obviously can't argue with you.... I don't know where in anything even remotely scientific that a 550% error can be acceptable, though.
Mike
For example, look at the This LInk In GH.
In the Anandtech Review, the difference between the pal6035 Alpha and a c-orb was a mere 2C, when the actual core temp difference is around 12C or higher.
Much more detailed information from experts like JohnCar, LaughingMan, Shoarthing, etc. in that thread, but the bottom line boils down to thermistor inaccuracy, and the impossibility of running a comparison on a socket A platform.
Of course, if you, like Tillman, believe that a 550% comparison error is perfectly fine, then I obviously can't argue with you.... I don't know where in anything even remotely scientific that a 550% error can be acceptable, though.
Mike
clumsum,
You just said it perfectly yourself. Temps changes should be represented by the thermistor underneath the socket, but they aren't. an actual 12C core temp difference is measured as 2C difference. It is this "compression" that prevents accurate comparisons to be done on socket A temps platforms.
The reason it only measure's a fraction of the change is explained in the thread in GH. Measuring from a secondary heat pathway is inaccurate at best.
Mike
You just said it perfectly yourself. Temps changes should be represented by the thermistor underneath the socket, but they aren't. an actual 12C core temp difference is measured as 2C difference. It is this "compression" that prevents accurate comparisons to be done on socket A temps platforms.
The reason it only measure's a fraction of the change is explained in the thread in GH. Measuring from a secondary heat pathway is inaccurate at best.
Mike
I can see your point..................but I think it is overstated, there are so many variables that you may be splitting straws, and calling them culverts......................just my thoughts..........however, I have been wrong before, maby even more than once. (Actually, much more than I like to admit!)
Thanks for the explanations..........and info.
I'm replacing my FOP32-1 with the Taisol CGK 742092, basically because I think it's a good cooler for the $$ and I really like the retention clip design........and I'm not oc'ing my 700/Athlon, at least for now.
Thanks for the explanations..........and info.
I'm replacing my FOP32-1 with the Taisol CGK 742092, basically because I think it's a good cooler for the $$ and I really like the retention clip design........and I'm not oc'ing my 700/Athlon, at least for now.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 23K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter