HS testing method suggestion

[Kakao]

The performance ratings page (http://www.overclockers.com/articles373/) compares Heatsinks coupled with some different fans. I think it is not very accurate since it is comparing also fans along HS. Why not also test the HS without fans since the test is done using a CPU die simulator without the risk to fry a real CPU? This way the comparison would be between HS alone.
What do you guys think?

[UnseenMenace]

While it is possibly more correct to test heatsinks in the manner you describe, the consumer does not buy them in this manner and as such it would have no real world relection or indication respective to the performance of the product in the way obtained by the consumer.

[arhines]

it would be nice to know if my ax-7 will keep a morgan duron alive for at least an hour or so though... or to know it it will passively cool an underclocked duron.
interesting idea, but I'm sure it would be lots of work to do this in addition.

[deeppow]

I suggested to a couple different on-line reviewers that they test using a reference fan (one 80mm and one 60mm) so we could see what the heat sinks did and the results wouldn't reflect some screaming fan that I'm never going to buy or run. All I got back was a couple of wisea** replies. :mad:

[Tiger]

Originally posted by Kakao
The performance ratings page (http://www.overclockers.com/articles373/) compares Heatsinks coupled with some different fans. I think it is not very accurate since it is comparing also fans along HS. Why not also test the HS without fans since the test is done using a CPU die simulator without the risk to fry a real CPU? This way the comparison would be between HS alone.
What do you guys think?

Lets analyse what you are saying here and then comment on the merits or demerits of such an approach.

You are saying that heat sinks should be tested based on their convective capability.

While convection is the base principle on removing heat from a heat sink in this particular context convection alone cannot supply sufficient air flow over the surface to effect sufficient cooling for the heat conducted from the cpu core. That is a fact and no heat sink will ever be able to supply sufficient passive (convective) cooling to cool one of these devices.
Since the above is a fact the manufacturer has had to design the heat sink around forced convection. Since the design incorporates features to maximise heat transfer by means of forced convection any tests must use forced convection as the basic tenant of the test.
In addition the manufacturer will probably have based the design around a minimum air flow velocity and a specific air flow pattern from a fan manufacturer. While increased velocity may improve performance a different flow pattern may alter the cooling performance of the heat sink.

[Kakao]

Thank you all for the good replies.

UnseenMenace,
Sure the consumer won't use a HS without a fan, at least not with a modern CPU, but which fan? The loudest, a compromise or the quietest? I'm using the 7V mod to keep the noise down. Many opt for the extreme cooling aproach, although I suspect that not for a long time. At the ratings page: "note that these rankings use air cooling heatsink fans as they are sold". Note that HS are sold with options of fans or without one.

Arhines,
Passive cooling was the idea that made me post this thread. If I knew that a certain CPU produces, say, 70 watts max heat, what theoretical HS would have the passive C/W necessary to cool it to a not dangerous 60C? Would it be possible using a 35 watts CPU?
it would be lots of work to do this in addition
I'm not sure. Why? The HS is already on the bench. The conditions are set.

Deeppow,
Reference fans. Yes, why not?

Tiger,
I like your points and your effort to clarify things, although I will discord:
You are saying that heat sinks should be tested based on their convective capability.
I'm saying that heat sinks should also be testet based on their convective capability.
convection alone cannot supply sufficient air flow over the surface to effect sufficient cooling for the heat conducted from the cpu core. That is a fact
Where did you get that fact from? Please point me to your sources. Conjecture alone is not enought to make facts.
no heat sink will ever be able to supply sufficient passive (convective) cooling to cool one of these devices
Thats a very definitive statement and I wouldn't be so courageous. In fact passive cooling has been used in CPUs before. Remenber the old 486? Those HS were small compared to today's standard.
In addition the manufacturer will probably have based the design around a minimum air flow velocity and a specific air flow pattern from a fan manufacturer. While increased velocity may improve performance a different flow pattern may alter the cooling performance of the heat sink.
I agree. One HS could be the best performer coupled with a fan and one of the worses without one (not very likely). I'm not proposing to give up with fan test, they are essential for a good analysis. I'm proposing one more element to consider in a HS performance.

[UnseenMenace]

Originally posted by Kakao
Thank you all for the good replies.

UnseenMenace,
Note that HS are sold with options of fans or without one.


Wow you sure, In the UK where i live, I have never been offered the opportunity to purchase a heatsink without a fan.
Another question is how do you suggest that we test heatsinks like the excellent Millennium Thermal Glaciator 2 which features the fan as a major part of the design, it is not just screwed on top... should we dismantle this heatsink and thus the very principles of its design in order to do the type of tests suggested here ?
And what of the special fan.. you can not just strap any fan to this heatsink, so a refrence fan would be pointless.
Do you honestly believe that in the situation of the Millennium Thermal Glaciator 2 that these tests would present correct information regarding its performance.
All these tests encourage would do is encourage the building of traditional style heatsinks rather than letting designers think outside of the norm to create quality and well designed products such as the Glaciator

[Kakao]

Wow you sure, In the UK where i live, I have never been offered the opportunity to purchase a heatsink without a fan
UnseenMenace,
In Brasil, where I live, I have very little options of heatsinks and fans, probably much fewer than you in the UK. But thanks to Internet we could buy from places like Newegg, very popular among the guys in this site, that sells heatsinks only (http://www.newegg.com/app/ViewProduct.asp?submit=manufactory&catalog=62&manufactory=1444&DEPA=1) or from 2CoolTek which also offers heatsinks without fan (http://2cooltek.safeshopper.com/34/340.htm?394).
All these tests encourage would do is encourage the building of traditional style heatsinks rather than letting designers think outside of the norm to create quality and well designed products such as the Glaciator
It is interesting how the way test designs influence the product design. It is real. This question you raise leads to a whole line of debate. Should we design tests to favor this or that trend? I won't go into it.

Edit: I'm pleased to deserve the attention of the very own forum moderator. Thank you.

[Tiger]

quote:
--------------------------------------------------------------------------------
convection alone cannot supply sufficient air flow over the surface to effect sufficient cooling for the heat conducted from the cpu core. That is a fact
--------------------------------------------------------------------------------


Where did you get that fact from? Please point me to your sources. Conjecture alone is not enought to make facts.


Based on the the AMD specs ( the heat sink must not exceed it's recommended mass) please design a heat sink that is capable of passively shedding 90W of heat. If it could be done, it would have been done already because noise pollution is becoming a serious factor, particularly in the work place. I don't need to prove it since there isn't a single manufacturer ( with all their thermal engineers) that has been able to achieve this. The closest thing that has been used is a heat pipe used in laptops but this is phase change therefore does not fit in this category.

quote:
--------------------------------------------------------------------------------
no heat sink will ever be able to supply sufficient passive (convective) cooling to cool one of these devices
--------------------------------------------------------------------------------


Thats a very definitive statement and I wouldn't be so courageous. In fact passive cooling has been used in CPUs before. Remenber the old 486? Those HS were small compared to today's standard.

Refer to my answer above.

With regard to the old 486 and previous cores the heat output was in the order of a few watts. We are talking about a core that is consuming anything up to 40amps of current.

If you want proof of this fact then do a search on this forum and find out how many K7 cores are now deceased because a fan was not plugged in. There will be hundreds. The fact is that most of them died in less than a minute.

[deeppow]

UnseenMenace, you argue for the status quo to allow the engineers the freedom to "get out of the box"? If I thought that was the case I might agree but it may more often allow marketing to put the highest cfm fan on the sink to make up the difference and test well. Just as importantly, non-standard testing allows the reviewer to screw it up.

There are clearly well-engineered heat sinks and companies with high product standards so that isn't my point.

While the Millennium Thermal Glaciator 2 may be an excellent heat sink, I see nothing to suggest a unique feature other than the fact the fan is built into the heat exchanger. So I decided to do a little research.

I pulled up a search engine and put in "Millennium Thermal Glaciator 2." Second hit was a review from TekSector (http://www.teksector.net/reviews/HSFs/millenium_thermal/glaciatorII/index.html). Read through the review and thought the reviewer made some excellent points related to the heat sink engineering. In particular he noted fan info as "it's a relatively quiet fan (36dba), whisking out a stream of 29 CFM @ 6000 RPM." OK, marketing didn't get to mess around with this design (so I eat some crow here, please pass the salt).

I then noted that up at the top of the page under "Latest Happenings at TekSector" is a review of the Thermalright AX7." Happens to be the same heat sink I'm running so I pull up the review --- same reviewer, great. But to my dismay, he places "the 80x38mm Delta FFB0812SHE ... pushes 68.5CFM of airflow with a whopping 48.5 dBA of screaming output" on the AX7. 49dBA? Not in my computer, I still like to listen to music with quite passages!!

Anyway, the test results place the AX7 in 3rd as follows:
1.) Crazy PC/Thermal Integration Dr. Thermal Extreme!
2.) Thermaltake Volcano 7+
3.) Thermalright AX-7
4.) Coolermaster HHC-001 Heat Pipe
5.) Global WIN CAKII 38
6.) Vantec CCK-6035D
7.) Arkua 7528
8.) Global WIN CAKII 68
9.) Global WIN TAK68
10.) Millennium Thermal Glaciator II
Now wait a minute, this isn't fair to the Millennium Thermal Glaciator II even if "my" AX7 did beat it.

So how unfair was the test?

Fully developed turbulent forced convection is a function of velocity to the 0.8 power (almost linear). Heat sinks also have a right angle turn of the air flow plus re-establishing the thermal boundary layer, etc. etc. Anyway, we need an estimate of the air velocity. We can make a WAG (wild as_ guess) at an answer as follows:
velocity ratio = (air velocity of AX7)/(air velocity of MTGII)
where air velocity = (volume flow rate) / area
so that the velocity ratio ~ 1.33 (Result neglects the fan's center hub.) The AX7 is moving its air about 33% faster than the MTGII. So lets say the AX7 had an advantage somewhere between 25% and 33%.

I'm impressed that the MTGII when so handcapped could come in 10th place!!

So who has the best heat sink? I don't know but I don't trust the TekSector ratings to tell me that answer (nor any other review ratings I've seen since they all typically do the same thing with fans). If I use the same fan - heat sink combo that they use in testing, their info does provide me a relative measure.

So what is John Q Public to do? Status quo doesn't work when testing may produce misleading results (note that this is my personal opinion). My original suggestion in this thread was to use reference fans to try and eliminate some of the inconsistent test results. That in fact was "short hand" for standardize testing. Without some normalization, John Q Public can and is being mislead.

I apologize to forum readers for the long message. :D

[UnseenMenace]

All valid questions and I think that this thread will only produce more questions than answers.

Such as If a refrence fan was used in the attempt to create more stable results. Would you produce results with the fan blowing air towards the heatsink or sucking air off ?
As I have obtained different performance levels with different heatsinks by moving the direction of airflow

[deeppow]

UnseenMenace, yes -- valid point(s). But standardized testing is present in a number of industries. Standarization can be an industry agreed to process. (The auto safety business is a clear exception were it was forced on them, at least in the US.) A reviewer or reviewing web site can use its own standard setup. The current results from Overclockers.com really is a reasonable start at this effort (http://www.overclockers.com/articles373/) with the standard heating die. Maybe you test their default fan setup (if one exists) plus the reference one.

More time consuming? Yes. The devil is in the details? Yes.

Overclockers.com and the others get to make up their own minds regarding their testing. Whatever the result in the heat sink testing arena, Overclockers.com has given me a forum in which to suggest my opinions. Plus the chance to interact with very knowable OCs like yourself and others! Its interesting and fun for a newbie. :D