CPU Low Profile CPU Cooler round up

IntoxicatedPuma

Customizer of Titles
Original poster
SFFn Staff
Feb 26, 2016
992
1,272
Are you planning to test larger footprint coolers like the Big Shuriken2, NH-L12(s) and AXP-100 that are still short enough for sff?

Sorry I missed this earlier, yeah I plan to test those as well. On the OP now I have the list of the upcoming coolers (next month or so) but I will start moving to larger ones after that (during June/July/August). Right now I have no plans do to anything over 130mm tall coolers so that will give us some 92mm towers but no 120mm towers. I think those should cover the NCase and others like Jonsbo U1 Plus.

then again an L9i and the Delta 92mm jet turbine would fit too...

You have my curiosity

92mm jet turbine

You have my attention
 

SaperPL

Master of Cramming
DR ZĄBER
Oct 17, 2017
472
882
That's quite nice resource you are making, but is there any chance that you'll be making a test run in real world scenario inside an SSF case?

In SFF case that doesn't have fans venting the case, the cooler itself will have to work within confined space to exchange the air. On open bench it is completely different scenario.

I understand though that choosing a case which would optimally fit most of the coolers while still being really small is pretty hard.
 

IntoxicatedPuma

Customizer of Titles
Original poster
SFFn Staff
Feb 26, 2016
992
1,272
It is different from a case, but I cannot account for all the variables of different case designs and also the time needed to disassemble such cases (and many would need the motherboard removed to change CPU cooler) would be too costly. Perhaps in the future during case reviews, I would take the top CPU fans that can fit in that case. Strictly speaking for the Cooler roundup I don't plan to test in different cases.

Another idea would be to simulate different scenarios and use a single top down cooler to test what the ideal space above a cooler would be for those height limited, and what position to orient the coolers. I don't have any plan to do this in the next 3-6 months however.
 

SaperPL

Master of Cramming
DR ZĄBER
Oct 17, 2017
472
882
Using a Midi/Mini tower case that would fit all those coolers but had only single (or no) front fan could be a good start.

That would at least include the fact of small cases not having too much induced airflow by fans as a contrary to open test bench where you have virtually unlimited access to fresh air.

To mitigate recycling of hot air, it would need a tunnel to the cover as most of SFF cases let coolers face the vents directly, but hard part here is what to do with tower coolers.
 

Tjo2012

Trash Compacter
Feb 6, 2018
34
11
@IntoxicatedPuma Any luck with the CGU-NC25TZ japanese titan cooler? Im just asking because i did not see it in the "upcoming testing" section. Sorry if im being a pest and i appreciate the hard work youve already convinced me to get an lp53 instead of an ID-40v3
 

IntoxicatedPuma

Customizer of Titles
Original poster
SFFn Staff
Feb 26, 2016
992
1,272
@IntoxicatedPuma Any luck with the CGU-NC25TZ japanese titan cooler? Im just asking because i did not see it in the "upcoming testing" section. Sorry if im being a pest and i appreciate the hard work youve already convinced me to get an lp53 instead of an ID-40v3

Sorry I should have put it in my upcoming list. I will try to do some quick tests this weekend.
 

IntoxicatedPuma

Customizer of Titles
Original poster
SFFn Staff
Feb 26, 2016
992
1,272
A little curious, but as mentioned in the C7 cu thread the Prime 95 may offer skewed numbers. Can I get feedback on the ideal software for testing?

Intel XTU
AIDA 64
Prime 95 v26.6
Prime 95 (newer than 26.6)
Others?

I noticed that when testing at 4.0ghz, Prime 95 (newer than 26.6) hits around 90-95w packaage, but Intel XTU and AIDA 64 will only hit about 65w. Mostly though, I'm concerned about which will be the most consistent and also applicable to real world scenarios. I realize that gaming won't push the CPU hard in all cases, so I am also considering other uses like Video Rendering which are much harder on CPU's.
 
  • Like
Reactions: VegetableStu

Tjo2012

Trash Compacter
Feb 6, 2018
34
11
A little curious, but as mentioned in the C7 cu thread the Prime 95 may offer skewed numbers. Can I get feedback on the ideal software for testing?

Intel XTU
AIDA 64
Prime 95 v26.6
Prime 95 (newer than 26.6)
Others?

I noticed that when testing at 4.0ghz, Prime 95 (newer than 26.6) hits around 90-95w packaage, but Intel XTU and AIDA 64 will only hit about 65w. Mostly though, I'm concerned about which will be the most consistent and also applicable to real world scenarios. I realize that gaming won't push the CPU hard in all cases, so I am also considering other uses like Video Rendering which are much harder on CPU's.


Find a blender render that takes the same time on your cpu as you would run on prime 95 or just a reasonable amount of time to get the temps stable. What sucks is that prime 95 gets the cpu throttling fast so finding other options is difficult but i think this is the "best" test

There should be plenty of sample downloads and its free, and its probably the most stressfulthing most general users will ever do in a realworld scenario.

Prime 95 is a torture test when we just want to know at what wattage with what cooler set starts hitting personally unacceptable temps. For example the c7 cu likely performs much better at 100% load than day to day use compared to the c7

We can compare power consumption with reviews and start forming a very reasonable set of performance expectations from our cpu/cooler combos.

If its too slow to do blender or adobe premier honestly just stick with prime 95 because it still gives results just not as accurately
 
Last edited:

IntoxicatedPuma

Customizer of Titles
Original poster
SFFn Staff
Feb 26, 2016
992
1,272
I'm noticing that Intel XTU stress test seems about the same on each test which seems reliable. I may stick with Prime 95 as you mentioned, but go to 26.6 if most people think that will be an accurate way to test.


EDIT:
Looking at some of my results, it would seem that regardless of the benchmark used, the temperatures are strongly tied to the wattage needed on that benchmark, as others have mentioned.

For example:
119.2w (Prime 95) 4.3ghz = FAIL
93.1W (Prime 95) 4.0ghz = 62C
80.5W (XTU) 4.3ghz = 55.5C
72.4W (Prime 95) 3.5ghz = 43C
65.2W (XTU) 4.0ghz = 43C
53.2W (XTU) 3.5ghz = 33.5C

I'm probably going to stick with Intel XTU simply because it's giving me more consistent wattage (5-10%) on each test, while Prime95 can tend to fluctuate by about 10-15%. This may also be due to the lower power XTU uses, but it sounds like it might be more "real world" accurate.
 
Last edited:
  • Like
Reactions: VegetableStu

dondan

Shrink Ray Wielder
DAN Cases
Feb 23, 2015
1,977
8,378
Prime95 is the best tool check overall system stability after oc. It creates a szenario that is the heaviest for the CPU (no normal software will reach it) so if your system can run a full 24h prime run ( default run through all tests) it is 99% stable. Only RAM can now be a problem (Memtest86 can check ram stability) .

Prime95 prior 27.x like 26.6 does not use AVX this is the reason why it generates less heat. AVX will be used only by a handful of software. Therefore Prime95 26.6 is the best tool for ITX systems where heat is your biggest problem. For max CPU temp you have to run 8k or 20k test with fft's in pace min. 15min. It takes some time before max temps are reached. For water cooling you have to run 30-45 minutes.

There is one interesting secreat in heatsink testing that nearly no ones know. For example if you buy 10 time the same heatsink they will perform different. I had L9i samples that are 5-8°C different. The reason is soldering and heatpipe quality. Normally heatpipes are sorted by their cooling power. Than you start mixing them so you will have the overall same cooling power, but variation of 10% is normal.
Also there is a variability of CPU plate quality and soldering of the heatsink to the fins.

Also the orientation of the heatsink will have a big influence on temperatures. Don't let the fluit inside your heatpipes work against gravity.
 
Last edited:

firewolfy

Master of Cramming
Nov 12, 2015
424
836
Great post Dan. Good point about testing system stability, and for the torture test using v26.6 to check max temperature.

I would add that my torture testing showed about a 3C increase in temp going from the large FFT (128k) setting to the small FFT (8k) that Dan mentions, so not a huge amount, but large FFT maximizes power consumption and tests more ram than small FFT, so I prefer it, and use it as my standard.

Some other experimenters use a blend setting (maybe Louqe?) which ramps loading up and down, and graphs showed the corresponding oscillation in temp and power. I don't like it since you have to figure out whether to use the avg temp or the peak, and what duration constitutes a peak, etc.

As far as cooler pipe orientation, I prime tested with a case in 3 orientations (normal vertical, rotated 90, flat) but didn't see any temp change at all. My cooler was the Big Shuriken2. Dan, do you have some details on temp changes you measured with orientation change?
 

firewolfy

Master of Cramming
Nov 12, 2015
424
836
I'm noticing that Intel XTU stress test seems about the same on each test which seems reliable. I may stick with Prime 95 as you mentioned, but go to 26.6 if most people think that will be an accurate way to test.


EDIT:
Looking at some of my results, it would seem that regardless of the benchmark used, the temperatures are strongly tied to the wattage needed on that benchmark, as others have mentioned.

For example:
119.2w (Prime 95) 4.3ghz = FAIL
93.1W (Prime 95) 4.0ghz = 62C
80.5W (XTU) 4.3ghz = 55.5C
72.4W (Prime 95) 3.5ghz = 43C
65.2W (XTU) 4.0ghz = 43C
53.2W (XTU) 3.5ghz = 33.5C

I'm probably going to stick with Intel XTU simply because it's giving me more consistent wattage (5-10%) on each test, while Prime95 can tend to fluctuate by about 10-15%. This may also be due to the lower power XTU uses, but it sounds like it might be more "real world" accurate.
Good point. Seems obvious now that you bring it up, lol. (the sustained wattage is probably the only thing that matters)
 

Tjo2012

Trash Compacter
Feb 6, 2018
34
11
Damn i hope the lg 1155 performs better with alternate fans.

But of a disapointment for the size and amount of materials atm
 

Tjo2012

Trash Compacter
Feb 6, 2018
34
11
I noticed you added some more tests @IntoxicatedPuma

Did you swap from prime 95 to intel xtu?

Also good work its appreciated

I just got zalman to send me a CNPS2X am4 bracket so im excited about that.

Edit* also in curious at which wattage/voltage people are most interested in. Seems like the high OC tests and low OC tests have some variability and are not linearly effective at disapating heat.