A Universal Formula To Rate The Performance of Any Cooling Solution

[Dr_Emmett_Brown]

What value does introducing this equation bring to us?

A meaningful metric that removes advertising hype.
A way to isolate sub-system components one a time to determine their overall impact on the total cooling. (change fan speeds and see a quantitative change, big or small, in the number)
A way to compare different radiators objectively.
A way to fine-tune your cooling solution with better TIM, measuring the effectiveness of de-lidding, etc.

What does a snapshot in time tell us about cooling capacity?

The short answer is a number that indicates how well your cooling solution is working at that moment. Bigger numbers are better, of course.
How much harder is your system working in the summer vs the winter? Compare your data over time by noting the changes in room temperature.

Results can be all over the map due to all of the variables not accounted for, right? What if I had the same hardware but better tim or block? Does this really show its potential? Or the current state only?

Now you're starting to get it, I think. Yes! Change ONE THING in your system, under same load same room temp, and you can observe how the INDIVIDUAL DIFFERENCE is better or worse than the prior one. I think you are thinking there needs to be a "TIM variable" and a "cold plate variable," but the beauty of it is, the COLLECTIVE effectiveness or ineffectiveness shows up in the OUTPUT TEMPERATURE. So every component of your cooling solution IS in the equation!

Better TIM = higher score than before.
Better cold plate = higher score.

So you can isolate and run your own tests and then put the best of ALL OF THEM in and have an even better system.

If you're still not sure about it, consider this.

My condenser with all that copper is so good, that whether I run 1 fan on it or 4, it won't change my HRQ number by more than 5%.

That means the CONDENSER is doing a bulk of the work at 200-205 watts.

If I ever get more cores/hotter chips, I know the fans will be contributing more once the system shows much more than the 200 watt being used.

 

[Blaylock]

A meaningful metric that removes advertising hype.
A way to isolate sub-system components one a time to determine their overall impact on the total cooling. (change fan speeds and see a quantitative change, big or small, in the number)
A way to compare different radiators objectively.
A way to fine-tune your cooling solution with better TIM, measuring the effectiveness of de-lidding, etc.




The short answer is a number that indicates how well your cooling solution is working at that moment. Bigger numbers are better, of course.
How much harder is your system working in the summer vs the winter? Compare your data over time by noting the changes in room temperature.



Now you're starting to get it, I think. Yes! Change ONE THING in your system, under same load same room temp, and you can observe how the INDIVIDUAL DIFFERENCE is better or worse than the prior one. I think you are thinking there needs to be a "TIM variable" and a "cold plate variable," but the beauty of it is, the COLLECTIVE effectiveness or ineffectiveness shows up in the OUTPUT TEMPERATURE. So every component of your cooling solution IS in the equation!

Better TIM = higher score than before.
Better cold plate = higher score.

So you can isolate and run your own tests and then put the best of ALL OF THEM in and have an even better system.

If you're still not sure about it, consider this.

My condenser with all that copper is so good, that whether I run 1 fan on it or 4, it won't change my HRQ number by more than 5%.

That means the CONDENSER is doing a bulk of the work at 200-205 watts.

If I ever get more cores/hotter chips, I know the fans will be contributing more once the system shows much more than the 200 watt being used.

The better number to use is measured in Celsius. I'm not sold.

 

[EarthDog]

So.... you've come up with a way to put a score to data? Ok. Cool, I guess! I just want the data.

Temperatures arent created equal, meaning what each cpu can handle. Some are limited to 72C, others 105C. Some have solder TIM, others crap TIM. It seems there are multiple variables not taken into account.

This feels like it is taking something black and white then adding and ambiguity to it? I dont know.

 

[Dr_Emmett_Brown]

Temperatures arent created equal, meaning what each cpu can handle. Some are limited to 72C, others 105C. Some have solder TIM, others crap TIM. It seems there are multiple variables not taken into account.

This is a law of thermodynamics.

It is ambivalent: it just spits out a quantifier that we can all use.

A CPU has a max temp of 72C before thermal shutdown. OK. It won't tax a cooling solution as hard as a Xeon putting on suntan lotion at 105C.

It's not the "formula's job" to get your cooling solution a high score. In fact, the whole notion of a "high score" is not germane to the formula.

Input: Watts, delta T, area from which the heat is generated, out comes a number.

The number tells the story of how hard the cooling solution is working under those exact conditions.

You do accept that one TIM might not be as good as another, but you don't connect the fact that this will result in different observed temperatures of the systems?

If ANY component is not as good as another, the system will run hotter.

And a hotter system will have a LOWER SCORE.

Understand? The equation accounts for this because the TEMPERATURE is in the equation.

This feels like it is taking something black and white then adding complexity and ambiguity to it.

I can't present it any more simply. Bring others to this thread whose opinion you trust. It must be read thoroughly and pondered, then it will be understood.

 

[Mr.Scott]

I can't present it any more simply. Bring others to this thread whose opinion you trust. It must be read thoroughly and pondered, then it will be understood.

IMO, you're just convoluting a principle here. I see no revolutionary formula.
Standing with Earth_dog and Blaylock here.

 

[Dr_Emmett_Brown]

IMO, you're just convoluting a principle here. I see no revolutionary formula.
Standing with Earth_dog and Blaylock here.

I'm not asking for a vote. I'm asking for those who truly understand to comment on it.

For example, earth dog is having trouble with the notion that the contribution of every individual component is accounted for in the equation. He has cited things about varying the TIMs and the "blocks" but he does not link them to the obvious fact that such inefficient parts will yield higher operating temperatures. And, of course, the temperature is in the equation. Once he understands that fully, a light bulb will appear over his head, and he will realize that different temps will yield different scores which means different performances as a function of better/worse components can be measured objectively.

Someone else, maybe him, I'd have to go back and look, did not think the CPU die area needed to be included.

A simple thought experiment should cure this misnomer.

Imagine your CPU, and therefore the cold plate, was the size of your thumbnail. It would be PRETTY DIFFICULT to design a cold plate to be useful, since presumably a great deal of heat is trying to be removed from such a small area.

Now imagine a CPU that is 16 square inches, 4 inches per side. That would be much easier to cool because there is so much contact area.

The die size is in the equation because that is the surface area of the heat generator.

Clearly everyone understands the role of delta T. The smaller the delta T, the better your cooling solution is.

Clearly everyone understands the role of wattage. The more, the harder it is to cool.

So for the life of me, I can't grasp what the group is unable to grasp.

 

[EarthDog]

So for the life of me, I can't grasp what the group is unable to grasp.

At this point I get it....but struggling to find the worth of this 'score'/method over, say, temperatures.

1. Temperatures are, for all intents and purposes, THE value to go by in the end, right? So why would users care about this relative value?
2. What does it tell me that a drop in temperatures would not already?
3. How do we use this data? If I score, 1% 'better' using your equation from a change, is that 1% better temperature-wise? How does this translate back to a tangible value for end-users?
4. Why wouldn't people just look at the temperature in the first place instead of having to go through this math?

If a user buys/builds a cooling loop, in the end, all they want to know is the temperature of the device they are cooling went lower. Will you kindly explain why a score relative to the wattage used and the size of the die matters in the end? Temperature accomplishes this already, yes?


Which cooling solution is better using these variables?

150W CPU, 90C CPU, 149mm2 Die size, 22C ambient

vs.

150W CPU, 70C CPU, 214mm Die size?, 22C ambient?

 

[Dr_Emmett_Brown]

150/90/22/149 = 14804 HRQ = better
150/70/22/214 = 14602 HRQ

I think I was IP banned. I can't reach the site from my computer. This is my cell phone reply, sparse as it is. Sorry this was so upsetting, I only wanted to help.

 

[EarthDog]

You haven't been IP banned/your account is fine AFAIK (You couldn't post from your phone if you were banned - PM me with what you are seeing if you are still having trouble...).

This is not upsetting. Just trying to see what we may be missing!

So 14,804 is better than 14602. This is 1.4% slower/less performance metric if I mathed right (plz ck!). That is the same cooling solution with two different CPUs (Q9650 and 8700K) pulling the same wattage (measured with software - different versions - for what little that is worth).

 

[Dr_Emmett_Brown]

One setup is doing slightly more heat removal than the other under those conditions.

The fact that the two numbers are so close means it rated the same cooling solution almost the same, despite a 20 degree difference in operating temperature.

Since I had no way of knowing this, doesn't that highlight the fact that the equation has merit?

If you just relied on temperatures alone, wouldn't you say it's doing worse on the 90C system than the 70C system? But as the equation shows, it's working nearly as hard on both.

 

[EarthDog]

If you just relied on temperatures alone, wouldn't you say it's doing worse on the 90C system than the 70C system? But as the equation shows, it's working nearly as hard on both.

To an untrained eye, sure. Otherwise, no. It's working the same on both. I see how this works, but, for me personally, just struggling to understand its overall usefulness versus temperature (when you compare like versus like).

 

[Dr_Emmett_Brown]

So you're classifying a 20-degree difference in temperature as "the same" because your eye is trained?

Yet two of my numbers with a 1% variance is somehow not useful?

To this I say: I've demonstrated the usefulness. An individual's understanding of it or lack of understanding doesn't change the fact that the formula is working as it should.

For all of the "untrained eyes" out there, this is objectively the way to go.

 

[EarthDog]

So you're classifying a 20-degree difference in temperature as "the same" because your eye is trained?

It is the same. It is still dissipating 150W load...your math shows a mere ~1.5% difference. I'd call that similar.

The properties of the cooling loop do not change from one load to another. What did change was the CPU below it and how it is able to get the thermals out to the block. We went over this earlier (remember the CPU block?).

If someone wants to use this equation to 'score' their cooling solution after they buy it, that is of course, fine. It works for that. That = placing an arbitrary value on their cooling performance based (in part) on the size of the die below the IHS. What do we compare this to? How does one know how well their system is working without a database of results (what about the unanswered questions in post 27)?

Me, I'll stick with temperatures as that data is easier to come by (zero math!).

So congrats! We know it 'works', the equation... but working and actually being useful are two different things. This is novel, but just isn't striking me as particularly useful for most enthusiasts or even those who are new (who would just look up reviews and see temperature/delta results).

GL!

 

[Dr_Emmett_Brown]

At this point I get it....but struggling to find the worth of this 'score'/method over, say, temperatures.

The score has 5 digits of granularity. Even a SLIGHT fractional difference in heat dissipation will show up more prominently than a 2-digit temperature reading.

1. Temperatures are, for all intents and purposes, THE value to go by in the end, right? So why would users care about this relative value?

In your example posted chronologically after this post, you showed a 90C temp and a 70C temp from two different CPUs using the same cooling solution.
There is NO WAY you can convince me that an outsider would "know" that with such a large variance in temperature, that each system was functionally removing the same amount of heat.

Use your own example. You say you want to use temperature to gauge performance. Guess what? 90 is hotter than 70, so by your own logic, 70 is better.

Use my formula, and you get a 1% difference. It sees them both as functionally the same.

2. What does it tell me that a drop in temperatures would not already?

A DROP in temperature would result in an INCREASE in the value. But, as the previous example highlights, you can't tell which is better when the temperatures are different from different CPUs.

3. How do we use this data? If I score, 1% 'better' using your equation from a change, is that 1% better temperature-wise? How does this translate back to a tangible value for end-users?

My one Thermosyphon is in the 21000 range. I think it's safe to say that is in a higher performance class. Condensers will outperform radiators. But by how much? This value helps answer that question. I've rated the CORSAIR H150i under different loads. Those numbers are posted here and you can compare them to your own. You seem stuck on the whole 1% thing, which is not the point. A 1% difference is negligible. And when the same exact system does report different numbers under different loads and different environmental temperatures, it just means in one case it is working harder than another case.

This can also be used to fine tune the cooling solution.

Do this: Misalign your water block intentionally. Or don't put the thermal paste on evenly or leave one spot bare. Then run the tests. What do you think will happen? It will run hotter, of course. Well what would you do without knowledge of the baseline HRQ number? You'd say, "Well, it's running hotter today. Hmmm." What would I do? I'd see the HRQ number is down 1000 points, and I'd start investigating each sub-system.

In this respect, it can be used to see how CHANGING ONE COMPONENT in a cooling loop can change the heat removal capability. I already discussed this in other posts.

4. Why wouldn't people just look at the temperature in the first place instead of having to go through this math?

You're still thinking too narrow. It is for comparing different systems, not watching the same system when it is 22C outside one day and 26C another.

If a user buys/builds a cooling loop, in the end, all they want to know is the temperature of the device they are cooling went lower. Will you kindly explain why a score relative to the wattage used and the size of the die matters in the end? Temperature accomplishes this already, yes?

I can't believe you just don't get it. I really can't.

 

[EarthDog]

The score has 5 digits of granularity. Even a SLIGHT fractional difference in heat dissipation will show up more prominently than a 2-digit temperature reading.

To what end though? What does that granularity offer? I've asked at least twice now what the score means? What does it mean when I score 16,000 versus 16,001, for example? Do I panic? How does this change my temperatures, exactly? Users want tangible, not arbitrary value that doesn't break back down to C.

In your example posted chronologically after this post, you showed a 90C temp and a 70C temp from two different CPUs using the same cooling solution. There is NO WAY you can convince me that an outsider would "know" that with such a large variance in temperature, that each system was functionally removing the same amount of heat.

I think we are in agreement here, are we not? Didn't I say that already? And really, who cares if they are functionally removing the same amount of heat?

Use your own example. You say you want to use temperature to gauge performance. Guess what? 90 is hotter than 70, so by your own logic, 70 is better.

Right. Those are two dramatically different CPUs, however. Using the same CPU/heat source, temperature is 100% valid. This is how reviews for heatsinks and water loops/AIOs are done.

You seem stuck on the whole 1% thing, which is not the point. A 1% difference is negligible. And when the same exact system does report different numbers under different loads and different environmental temperatures, it just means in one case it is working harder than another case.

Im not stuck on 1% of anything. That is just a dataset and example for a talking point... not The Gospel.

Tell me, if I set up a loop using the same exact hardware/system and temperatures are different, does this mean the cooling system is working harder/easier or does it mean another variable has changed? You keep saying the cooling system is working harder, but, the cooling system works the same every time (assuming all other things equal).

A DROP in temperature would result in an INCREASE in the value. But, as the previous example highlights, you can't tell which is better when the temperatures are different from different CPUs.

Right, x2. But who uses different CPUs/heatsources for a review? Nobody best practices. If you look at most any reputable review, the information presented is in C, be it absolute or a delta.

In this respect, it can be used to see how CHANGING ONE COMPONENT in a cooling loop can change the heat removal capability. I already discussed this in other posts.

Temperature already does this. I've said this too...

You're still thinking too narrow. It is for comparing different systems, not watching the same system when it is 22C outside one day and 26C another.

I don't recall mentioning anything about different ambient temperatures.

In this respect, it can be used to see how CHANGING ONE COMPONENT in a cooling loop can change the heat removal capability.

Temperatures can do the same thing.... when you are comparing it as you compare this one... with like hardware/heatsource.


I think I get it... or if I don't, myself and a couple others in chat are all struggling to find where this is the "Holy Grail" of objective measurement performance is. Temperatures tell us what we all want to know already... be it a whole loop or making a single change and seeing what the difference is. Temperature does the job when comparing cooling on the same system.


Sorry, bud... but not sure if I am to apologize because I 'don't get it' or because this isn't quite what you want it to be. I'll pipe down and see if there is something that flips the light on, however.

Thanks again for taking the time to try and explain this.

 

[mackerel]

I only quickly skimmed this thread, after seeing the other one calling for results. It has interested me enough I'll give it a go just to see how scores end up, and if something possibly meaningful might come out of it. The normalisation for die area is interesting. I have similar coolers on Intel and AMD and it will be interesting to see how that comes out.

 

[Dr_Emmett_Brown]

So here a the 3-stage vapor phase change "Thermosyphon Type A" video with 2 loud, high speed fans. I zoom in at the end so you can see the vapor boiling.

Ummm... wait how do we upload short videos, or don't we?

 

[Blaylock]

You need to go through Youtube.

Once it's uploaded to Youtube the type [yvid]enter vid address[/yvid]. EXAMPLE [yvid]LXwKPjwk5HA[/yvid]

and the result will be...

 

[Blaylock]

Based on your results thread the Noctua NH-D14 and NH-D15 had a higher cooling potential than your thermalsyphon? Is this right?

 

[Dr_Emmett_Brown]

Actually, I worked out via retrograde analysis that the Noctura numbers are incorrect.

1. He used the publishes TDP number for the CPU rather than the actual power drawn.
2. He reported the Case T instead of the Core T.
3. Per the Intel specs for the 6700K, at the TDP of 90 watts, the case T will be 64 C, resulting in a core T of 100 C. This works out to an internal thermal resistance of 0.483 deg C-cm2/watt for the die and cover (which is lower than that for a Core i9 9900k, at 0.525 deg C-cm2/watt).
4. This implies at 90W, the 6700K chip would have a core T 36 C hotter than the case T.
5. If he core T was 56 C at 90 watts, the case T (at the base of the heat sink) would have to be 21 C, which is cooler than the air T (not physically possible).

So something is wrong with the mackeral numbers.

If it was the case T, the core T would be 92 C, and HRQ = 10,540 (not 21,700).