Looking for suggestions for heatsinks to review

[b1lk1]

Just using a heat source does not tell you how a heatsink will work inside a case on an actual working PC. We don't believe in that type of testing and it introduces far too many variables. In theory it is not a bad idea, but in practice it just isn't realistic.

 

[Optix]

I like heatsinks reviewed in real world applications.

 

[Psi*]

This took ... 1 minute ... to find.

I suppose the out of the box base plate roughness measurement would also be out of the question.

As far as an actual case ... which actual case & what actual case fans ... etc.

 

[b1lk1]

Believe me, I understand WHY you are asking these questions, but this not only adds greatly to the cost of building a setup since we currently use parts given to us by manufacturers to build test machines and there is no manufacturer for a setup like this, there is just no need to get that critical. If you understand what temperatures we get with out setup, it should take little thought to figure out how your different system will run the same heatsink. In fact, that TEC powered heating device will not heat up the same as a CPU would so if you ask me it adds yet another variable instead of taking variables away.

 

[Agent L]

Just using a heat source does not tell you how a heatsink will work inside a case on an actual working PC.

Exactly, because the point of the experiment is to test heatsink, not your case.

We don't believe in that type of testing and it introduces far too many variables.

Now you're trying to tell us that white is black. Are you a politician or what?
Here are some random variables present in real world setup:

1. unknown and uncontrollable power output of a real CPU
2. low precision of on-die thermal diode
3. matching diode output to program used to read values
4. airflow restrictions imposed by particular case, PSU and motherboard setup
5. external airflow introduced by PSU and case fans
6. air temperature affected by other components preheating it.
7. heat removed via motherboard conduction

Synthetic heater is a way to get rid of those variables.

In theory it is not a bad idea, but in practice it just isn't realistic.

It's as realistic as it can get, if one wants to test CPU heatsinks.
If you want to test entire rigs, sure, your methodology is superior. Just please, don't call it heatsink test, because it isn't one.

I can design a setup where passive HR-01 will outperform top active cooler. It's very simple, just block active cooler's fan with PSU wall and let case fans do sideways flow. What would it prove? That it's one very special rig, nothing more.
Less extreme example: Does your case have so called "prescott duct" and fan? Mine has, so tower coolers in my case will never be a match for a top-down cooler force fed with fresh air by a dedicated 120mm fan.

I know that building a proper test stand takes time, effort and money. I understand that you're unwilling to do it, and that's fine. It's your test, not mine. I'm not trying to force you into anything.
Just don't try to tell us that your method takes away randomness, when it introduces it. Because that's plain lie.

Only thing that synthetic might not replicate exactly is the uneven heat distribution pattern of the CPU (which is questionable difference due to heat spreaders in modern CPUs). However, it's very simple to have many differently sized heaters (as heater die is one of the cheapest components) AND to test if heater size do actually have any influence on performance.

We can figure out, you say. How tests done on i7 relate to, let's say, Athlon 2650e? (15W TDP) How test results from forward-to-backward flow relate to top-down case?
Well, I can't figure out. Call me stupid.

 

[Optix]

Simmer...

The test method used by Techgage is more indicative of what the average user will run into anyway. Sure, not everybody has the same setup but most enthusiasts that I know run a tower cooler with at least a front intake fan and a rear exhaust fan.

 

[Kougar]

This took ... 1 minute ... to find.

I suppose the out of the box base plate roughness measurement would also be out of the question.

As far as an actual case ... which actual case & what actual case fans ... etc.

Interesting, I was never aware of their synthetic test mounting system! My only question is how sure are they that 150W is sufficient?

The difference between an overclocked CPU at idle versus load by itself can add an additional 200W of power consumption. While all of that certainly isn't going to the processor alone, we can't forget the CPU is already consuming power when at idle in addition to the 200W... I think they need at least a 200W heat test to fully represent a medium overclock of a quadcore/hexcore CPU. Most people are not going to be paying $70-100 for the best coolers just to run them at stock or small overclocks.

 

[Tharic-Nar]

A synthetic benchmark is just that, synthetic, it's like battery life on laptops. Sure, we can run tests under a best case scenario, but they wouldn't hold up to real world use. Like testing a toaster as a radiator to see how much heat it produces, they're meant to be used a specific way and thus should be tested that way.

Testing a heatsink under lab conditions just means it works in the lab. Get some dust on it, mount it horizontally instead of vertically due to a height restriction on the PSU, etc, etc, and you end up with something very different. Maybe the mounting mechanism doesn't like AM3 sockets, maybe it only works with a specific kind of thermal paste. Can you test the backplate mounting accurately on a synthetic rig? If a product only works in the lab and not on the bench, then the product is worthless. If it can't handle 'real world' mistakes, then it has a design issue to contend with. You may end up with some variance in the final details due to 'human error', but thats the point, they're never going to be mounted optimally, and if they can't handle that variance, thats their problem.

Synthetic tests are an addition to regular testing, you can not replace one with the other, both are needed for an all round view. Synth gives a comparative baseline and real world testing gives you usability/'fit-for-purpose' results.

 

[b1lk1]

Thread closed. This was not opened up as a discussion of our testing methodology, I was looking for insight as to what we will try to get in for testing. While I know I cannot please everyone, the fact that this is turning hostile forces me to close this. Feel free to start a discussion thread if you like, but please keep the hostility to a minimum.

One very important thing I feel I must state. We do not get paid by Techgage or any manufacturer to do these reviews and testing. I do this because I enjoy playing with the latest and greatest parts. I am a stay at home dad with a wife and two kids. I cannot take money out of my budget to build a custom heat sink testing platform or buy highly specialized equipment for such testing.

I will go on record, right here, right now, stating that I do not believe in such ultra-synthetic testing as using a platform as was linked to. Noone has to agree with me, but that type of testing is NOT real world in any way, shape or form. Most of the world is not looking for that last 1C temperature drop so it is just no applicable anyways. Again, you do not have to agree with me, but that is the way I feel about it.

Again, feel free to open up a thread discussing our testing methodology if you like. We are not afraid of criticism. We have nothing to hide. All we want to do is test hardware in as close to a real world application as possible.

Thank you.