Air cooling tests and Delta Temperatures

[trdonja]

WHAT'S THIS ABOUT

Hi, thanks for checking out this topic. It's my first post, but I actually created an account to ask advice on my next PC. I guess that will be my second post then.

First about my motivation. Pure curiosity, not meant as a criticism. I sent a question as a response to the Noctua-NZXT video, but since I haven’t received a reply I just decided to do the experiment myself and see if my theory holds water. So what was my theory?

Most testing sites use delta temperatures as values that can be compared among themselves to get valid testing results. I however had my doubts whether same CPU cooler really performs with same efficiency in different ambient temperature. I thought that in that case NZXT cooler could have been the better cooler but was unfortunate to “compete” second. I have to say that personally I couldn’t care less which one performs better, I own neither of them and for me they both perform the same, regardless which one gets the small advantage.

So this was my experiment. The hardware is not really important, old AMD with some bad cooler. I am going to give more information than somebody would think is needed, just to show the conditions better and to give better foundation for result interpretation.

TEST 1

Ambient temperature before the start of test 1: 20.8 C

Core 1 temperature before the start of test 1 (stabilized, was not rising) : 51 C

Core 2 temperature before the start of test 1 (stabilized, was not rising) : 51 C

Test for 35 minutes: Prime95

Ambient Temperature at the end of test 1: 21 C

Core 1 temperature at the end of test 1: 81 C

Core 2 temperature at the end of test 1: 82 C

Delta temperature cores (start-end): 30.5 degrees

Delta temperature cores (end)-ambient (average): 60.6 degrees <-- this is the one that various review sites use on their charts

After that I let the PC cool down a bit, I was not using it. Both cores dropped to 52 C. I turned on the heating in the apartment, still not touching the PC. As the room temperature was rising, so did the CPU temperature. At one point I decided, that’s it, let’s repeat the test.

TEST 2

Ambient temperature before the start of test 2: 26.3 C

Core 1 temperature before the start of test 2 (stabilized, was not rising) : 55 C

Core 2 temperature before the start of test 2 (stabilized, was not rising) : 56 C

Test for 35 minutes: Prime95

Ambient Temperature at the end of test 2: 26.6 C

Core 1 temperature at the end of test 2: 89 C

Core 2 temperature at the end of test 2: 91 C

Delta temperature cores (start-end): 34.5 degrees

Delta temperature cores (end)-ambient (average): 63.6 degrees

Temperatures 5 minutes after test was ended:

Core 1: 58

Core 2: 59

Temperatures 9 minutes after test was ended:

Core 1: 56

Core 2: 56

Temperatures 12 minutes after test was ended:

Core 1: 55

Core 2: 56

Temperatures 15 minutes after test was ended:

Core 1: 55

Core 2: 56

I let the ambient temperature drop again (by turning off the heating) and started third, last part of experiment. Again, CPU temperatures started dropping when room temperatures was going down.

TEST 3

Ambient temperature before the start of test 3: 22.6 C

Core 1 temperature before the start of test 3 (stabilized) : 52 C

Core 2 temperature before the start of test 3 (stabilized) : 52 C

Test for 35 minutes: Prime95

Ambient Temperature at the end of test 2: 22.4 C

Core 1 temperature at the end of test 2: 84 C

Core 2 temperature at the end of test 2: 85 C

Delta temperature cores (start-end) : 32.5 degrees

Delta temperature cores (end)-ambient (average): 62 degrees

Temperatures 5 minutes after test was ended:

Core 1: 55

Core 2: 56

Temperatures 9 minutes after test was ended:

Core 1: 54

Core 2: 54

Temperatures 12 minutes after test was ended:

Core 1: 53

Core 2: 53

Temperatures 20 minutes after test was ended:

Core 1: 53

Core 2: 52

Ambient: 22.3

-- Edit: done another test, just to confirm consistent results:

Ambient: 21.3

Cores: 51 and 52

Prime95 for 35 minutes

Cores 82 and 82

End of Edit --

CONCLUSION

Ok, so just to summarize:

- Same PC, same test, only ambient temperature changing

- Ambient temperature 20.9 C had CPU running at delta temperature 60.6 degrees

- Ambient temperature 26.4 C had CPU running at delta temperature 63.6 degrees

- Ambient temperature 22.5 C had CPU running at delta temperature 62.0 degrees

The theoretical explanation is a bit of a a guesstimate, but it makes sense to me. If the fans would intake air from outside, that would always have the same temperature, then it would make sense that efficiency of the cooler is relatively similar at different room temperatures. However since the cooler takes in the same hotter air, it can’t cool the CPU as efficiently. Hot air is not able to take as much heat away from the CPU as cold air can.

I’ll be glad to hear your opinions and interpretations. I have to say that the result was exactly what I was expecting, delta temperatures can’t be compared, it has to be tested at same temperature. I will repeat one more time that the only reason I did this, is to satisfy my curiosity. I've been following this YT channel for a long time, because the tests are the best on the internet. It's just these delta temperatures that bothered me so far...

 

[sheroo]

This makes interesting reading...

http://www.rwlabs.com/category.php?page=cpu_heatsink_testing

 

[trdonja]

It is interesting reading, especially the part about controlled temperature with the lamps.

The air surrounding a heatsink plays as much of a part in the cooling devices performance as unit construction does, in some cases more so. This lack of control makes for careless evaluations since the performance can change not only day to day but even hour to hour. If you are trying to determine performance between multiple products then the old way is as accurate as rolling the dice.

On the other hand it's interesting that DH14 has the same delta no matter what the ambient temperature is:

http://www.overclock3d.net/reviews/cases_cooling/coolermaster_hyper_212_evo_review/4

Ambient 23.3, Delta 40.2

http://www.overclock3d.net/reviews/cases_cooling/nzxt_havik_vs_noctua_nh-d14_review/1

Ambient 19.5, Delta 40.25

Not sure how that is possible.

 

[sheroo]

If you were to do it completely scientifically, you'd have to remove all of the variables other than the heatsink itself. However for coolers such as the Corsair closed loop, or custom water cooling loops this might prove difficult.

Then there is the case when 2 coolers perform equally well but the noise levels are totally different. You would then need a setup that would be configured for a constant noise level, but then you would need an anechoic chamber etc etc.

Its very difficult. There was a guy on here a while ago that was talking about the number of significant figures that you can use for measurements as well - which was also a valid point. (Tom's ambient is measured to 1 decimal point while HWMonitor has no decimal places). Then to conclude that one heatsink is 0.5C better than another is, well how shall I say - not completely scientific.

Its all about time and resources. I would imagine doing heatsink testing is a real ball ache, and lets face it, Tom actually makes it quite entertaining. Whilst it may not be 100% accurate and scientific, it does give you a feel as to how good the coolers are.

Lets face it, whats a couple of degrees amongst friends.

 

[Timbaloo]

Was it a cooler using heatpipes? Heatpipes are somewhat limited in there effective temperature range. Heatpipes use phase changes to transport heat effectively, so they rely on vaporisation (hot side) and condensation (cold side). Now if the temperature (around the cooler's fins/cold side of the heapipe) is higher than the condensation temperature, then the main effect is lost for heat transport, and the cooler will obviously perform less effective.

 

[trdonja]

Timbaloo, it was AMD stock cooler.

Sheroo, good points, I agree. As I said, both coolers look pretty much the same to me when it comes to performance.