Home grown 120mm Thermal experiment/Showdown

[decripple]

I am an engineering student and an enthusiastic and curious one at that. I have thrown a bunch of money at buying recommended 120mm fans of varying speeds in the hopes of doing a cooling comparison. Also purchased was several brands of thermal compounds (TIMs) to test out in the same setup. For me, it is trying to also find what fans I will recommend for price/performance to all the people I do builds and upgrades for. So here is what is happening starting tomorrow afternoon:

 

I have a test bench and 7 pairs of medium - extreme speed 120mm fans that will be swapped on and off of my Thermaltake Water 2.0 Pro AIO unit. I have 5 TIMs and 1 "Extreme performance" thermal pad to test out

 

Method/Test Setup:

Each setup of TIM, pads, and fans will be run for 2 runs at least 20 minutes of stress testing looking for a thermal ceiling lasting at least 3 minutes. Temperatures will be recorded using Asus Sensor Recorder and OCCT small data set which I will be using as the stress test method to generate maximum heat.

 

For the regular thermal paste style materials I will be doing 2 "mounts" of each paste to get some sort of average. This means I will need basically a 12-16 hour day to test the non-curing pastes and for the curing paste(s?) I will run last again with 2 runs and this will take a few days of curing for each run. Or I might be lazy and ignore them (Arctic Silver 5) because I am impatient :]

 

With the thermal pads I can only manage 1 "mount" of the waterblock for each pad. This means 2 mounts total as well since I have 2 pads of the same material (they are way too spendy to do any more)

 

Fans will be ran in single fan "pull" exhaust and dual fan "puch/pull" setups. All fans will be ran at 100% RPM levels with the exception of the 2 fans that are super high speed which will also be run at a lower ~1700-2200 RPM level. I do this because all of these fans can be manipulated to run slower speeds using different methods of fan control. Also, I am looking for fans that will work for me with benchmarking and super high OCs on water.

 

The water block pump will be at 100% RPM to make it as consistent as possible

 

Hardware:

 

The hardware used in the testbench is as follows:

AMD FX-8350 CPU which will be running at 4.8GHz and 1.475V vcore (maybe higher but we shall see what this thing can handle) 

Corsair 2x4GB 1866 CL9 RAM 

Asus Crosshair V Formula-Z mobo 

OEM AMD R9 270 GPU 

Tt Water 2.0 Pro CPU cooler with assorted fans and TIM 

Samsung 840 250GB boot drive 

WD SCORPIO black 500GB storage drive 

Cooler Master Haf XB test bench with high speed Prolimatech 140mm aluminum fans in front to aid cooling

 

Here are the TIMs:

Noctua NT-H1 

Prolimatech PK-3 

Arctic Cooling MX-4 

Gelid GC-Extreme

XSPC K2

Cooler Master cheap stuff that comes with their heatsinks (assuming it is the same as their retail Thermal Fusion)

FujiPoly Ultra Extreme .5mm thickness 50x60mm sheet thermal pad

and Arctic Silver 5 if I get to it (doubtful)

 

Here is a list of the fans to run the gauntlet:

Corsair SP120 Performance editions

Scythe Grand Flex 2000RPM

Scythe Gentle Typhoon 4250RPM & running a resistor at 7-9V or something like that

Enermax TB Vegas RED LED @ 1800RPM

Enermax Twister Pressure also 1800RPM

Silverstone FN123 low profile 15mm @ 2000RPM

***Under dog/performance value*** Y.S. Tech FD121225HB 2600RPM Double ball bearing & running a resistor at 7-9V or something like that

***Underdog/value*** Masscool BLD-12025V1R RED LED ball bearing 1500RPM 

 

Ironically I don't have the stock Thermaltake fans anymore but since you can't easily buy them retail I don't think the exclusion is a big deal

 

 

 

Here are the fans next to the Cooler Master Haf XB I am using for the bench.

 

 

Testing cycles begin tomorrow afternoon! 

 

As Borat said "I am very excite to start my reporting"

[dadasmithywinkle]

Looks like it's gonna be awesome

[SuperPug]

Cool! followed...

[soaringchicken]

cool project, might borrow the findings in my own recommendations Thanks

[drewjn]

Very interesting. It is topics like this that make registering and following the linustechtips forum worth it. Hope your experiment goes without a hitch.

[decripple]

I should note that I wasn't confortable going above 80C so I didn't even finish single fan tests unless they could manage to stay in the mid 70s. Only the last 2 fans were able to do this. Also, I didn't run the high speed fans at lower speeds because I couldn't create a consistent speed and read the RPM results. Also, I was running out of time so I figured 100% speed for all of them is fair. I might run some of the better fans in another suite of tests later but for now here are the results. 

 

Fan results: 

Silverstone FN123: 78C 

Masscool: 77C 

Enermax Vegas: 75C 

Enermax UCTP12P: 75C 

Scythe Grandflex 2000RPM: 72C 

Corsair SP120 Performance: 75C 

Single YS Tech fan: 73C 

Dual YS Tech fans: 69C 

Single Gentle Typhoon 4250RPM: 68C 

Dual Gentle Typhoon 4250RPM: 66C 

 

These temps are all very high IMO, but to be fair this is a very hot CPU. As a reference point to show how brutal Small FFT OCCT is I played BF4 for 2+ hours in 64 player conquest with the Silverstone FN123s on it and never exceeded 50C. 

 

Another thing to note is that fan performance tracked directly with fan noise with the exception of the Silverstone slim fans which were slightly louder than the Masscool fans. 

 

Tomorrow I will be starting the TIM comparison because it takes much more time.

 

I think after the TIM results I will try out my Eisberg pump and my Alphacool 240mm UT60 radiator with the slower turning fans to see what temps I can get with 100% more surface area and 2-3x more tubing volume.

[rentaspoon]

I should note that I wasn't confortable going above 80C so I didn't even finish single fan tests unless they could manage to stay in the mid 70s. Only the last 2 fans were able to do this. Also, I didn't run the high speed fans at lower speeds because I couldn't create a consistent speed and read the RPM results. Also, I was running out of time so I figured 100% speed for all of them is fair. I might run some of the better fans in another suite of tests later but for now here are the results.

Fan results:

Silverstone FN123: 78C

Masscool: 77C

Enermax Vegas: 75C

Enermax UCTP12P: 75C

Scythe Grandflex 2000RPM: 72C

Corsair SP120 Performance: 75C

Single YS Tech fan: 73C

Dual YS Tech fans: 69C

Single Gentle Typhoon 4250RPM: 68C

Dual Gentle Typhoon 4250RPM: 66C

These temps are all very high IMO, but to be fair this is a very hot CPU. As a reference point to show how brutal Small FFT OCCT is I played BF4 for 2+ hours in 64 player conquest with the Silverstone FN123s on it and never exceeded 50C.

Another thing to note is that fan performance tracked directly with fan noise with the exception of the Silverstone slim fans which were slightly louder than the Masscool fans.

Tomorrow I will be starting the TIM comparison because it takes much more time.

I think after the TIM results I will try out my Eisberg pump and my Alphacool 240mm UT60 radiator with the slower turning fans to see what temps I can get with 100% more surface area and 2-3x more tubing volume.

Yay Silverstones! If your not happy going over 80 have you thought of dropping down to 4.6 or 4.4? Also I would have a reference fan and stick with it, test the paste then redo the fans at 100% then at what you would think is silent.

[spwath]

Yay Silverstones! If your not happy going over 80 have you thought of dropping down to 4.6 or 4.4? Also I would have a reference fan and stick with it, test the paste then redo the fans at 100% then at what you would think is silent.

I like Silverstone fans, specifically ap123

[SuperPug]

specifically ap123

 

Would have never guessed that *ehm avatar ehm*

[Ghost]

You should definitely add Liquid Ultra to your test. It will win hands down.

 

You need to test it last as it will bond to your CPU and heatsink though (you need to sand it off), however, there is no need for re-application, ever.

[decripple]

You should definitely add Liquid Ultra to your test. It will win hands down.

 

You need to test it last as it will bond to your CPU and heatsink though (you need to sand it off), however, there is no need for re-application, ever.

 

Considering the massive down sides that has I will not be doing that. Too much extra labor and hassle. Also, I would have to buy it. Which costs money and time waiting for it to arrive. Maybe in the future though.

[Ghost]

Considering the massive down sides that has I will not be doing that. Too much extra labor and hassle. Also, I would have to buy it. Which costs money and time waiting for it to arrive. Maybe in the future though.

Only downside is that you void your CPU warranty. Sanding it off takes seconds and in fact, sanding down heat-sinks and CPUs so the surface is flatter is a method of improving the temps. It's called lapping.

 

Honestly, besides the cost, which isn't really that bad when looking at the price of your whole rig, and that this will last you 10-20 applications and will never dry out. I think everyone who is serious about CPU cooling should use it. Application is really not that difficult and it's undeniably the best performer after Indigo Xtreme which will actually destroy your CPU and heatsink and is a huge pain to apply

[decripple]

Do you have any sources for this claim of the best compound? It is demonized quite often in many places I have looked for results (forums)

 

Also, how about the Coollaboratory Liquid Metal pads by comparison? I am thinking of considering your offer to try this stuff out... we shall see what evidence you can provide for it's case before I pony up the cash to try this stuff out.

[rentaspoon]

Do you have any sources for this claim of the best compound? It is demonized quite often in many places I have looked for results (forums)

http://forums.overclockers.co.uk/showthread.php?t=18416533

Don't just look at the bar graphs though read the whole thing

@Ghost voiding my warranty is a big thing for me, why use that when for a few degrees more I can have something that doesnt

[decripple]

http://forums.overclockers.co.uk/showthread.php?t=18416533

Don't just look at the bar graphs though read the whole thing

@Ghost voiding my warranty is a big thing for me, why use that when for a few degrees more I can have something that doesnt

Soooo....It isn't that great? Lots of complaints there and 2.75C difference from Gelid GC-Extreme so jury is still out. I will end up checking it out I think as the very last paste, but not this week.

[Ghost]

http://forums.overclockers.co.uk/showthread.php?t=18416533

Don't just look at the bar graphs though read the whole thing

@Ghost voiding my warranty is a big thing for me, why use that when for a few degrees more I can have something that doesnt

Bear in mind that is Liquid Pro, not Liquid Ultra which is much easier to work with (it's far easier to apply and just wipes off in most cases) and performs better/equally in most tests.

 

forums.anandtech.com/showthread.php?t=2297140

http://www.tomshardware.com/reviews/thermal-paste-performance-benchmark,3616-18.html

[rentaspoon]

Bear in mind that is Liquid Pro, not Liquid Ultra which is much easier to work with (it's far easier to apply and just wipes off in most cases) and performs better/equally in most tests.

forums.anandtech.com/showthread.php?t=2297140

http://www.tomshardware.com/reviews/thermal-paste-performance-benchmark,3616-18.html

Good spot. Though I don't think the benefits out way the risks (IMO) but I suppose you could call it enthusiast grade product

Soooo....It isn't that great? Lots of complaints there and 2.75C difference from Gelid GC-Extreme so jury is still out. I will end up checking it out I think as the very last paste, but not this week.

IMO Not really, read the article Ghost listed, they go through possible issues, I've got an old single core athlon I intended to use some on first but decided what was the point as other pastes didn't have the issues, cost less and only run a little hotter

[Ghost]

Good spot. Though I don't think the benefits out way the risks (IMO) but I suppose you could call it enthusiast grade product

IMO Not really, read the article Ghost listed, they go through possible issues, I've got an old single core athlon I intended to use some on first but decided what was the point as other pastes didn't have the issues, cost less and only run a little hotter

At £2 per application the price isn't really too bad. Of course, this isn't for someone running a stock heatsink, or even a cheaper heatsink. This is more for people who are water-cooling and spending so much money that the extra £2 won't even take up 1% of their budget.

[decripple]

Finally the experiment is over! Here are the screen shots of my exact settings while running the stress test and my overclock settings:

 

 

Digi+ Settings for power regulation stuff:

 

 

TIM application example of the rice grain/dot method:

 

 

Power Use from the wall with CPU at above settings and R9 270 @1100111010/5500 1.210 vcore 115% power:

Yes that is in watts.... Keep in mind PSU is 80+ Bronze rated CX750M PSU:

 

 

Other Random Test bench pictures taken through the top & side mesh of the HAF XB:

 

 

 

Results: 

Arctic Cooling MX-2: 

73C 72C 

XSPC K2: 

75C 74C 

Arctic Silver Ceramique 2: 

74C 80C 

Cooler Master Black tube (Eiscreme): 

74C 75C 

Cooler Master Pink Tube (212 Evo paste): 

73C 74C 

Prolimatech PK-3: 

73C 72C 

AC MX-4: 

72C 74C 

Noctua NT-H1: 

74C 72C 

Gelid GC-Extreme: 

70C 72C 

 

FujiPoly (core frequency and vcore voltage + average temperature of 2 runs): 

3.8GHz @1.35V: 56C 

Stock Settings = 

4.0GHz Turbo off 1.375V: 62C 

4.4GHz 1.4375V: 68C 

4.6GHz 1.475V: 79C 

 

So pretty much everything performed as expected with the Gelid being a solid 1-2C improvement over the other pastes. This is a very small lead as my test bench ambient was within 1 degree celcius ambient environment. Realistically all the pastes are good enough to handle overclocks, but the MX-2 and GCE where easier to clean and apply IMO. 

 

The Ceramique is get what you paid for. It was terrible to clean and apply as it was very thick and viscous. All other pastes were similar in ease of use and perform very similar, too. 

 

Lastly the Fujipoly pads were cut down to about an 1/8 inch beyond the size of the CPU heatspreader to minimize waste and make mounting the Tt AIO easier. They were the most consistent bringing in identical load temps at their settings. You can see how I gradually stepped up the load until I reached my "safe" limit for Temps which I arbitrarily chose as 80C. 

 

Price to performance crown definitely goes to the Arctic Cooling MX-2 which comes in a pretty good sized 4gram tube for only 8 bucks and allows for a solid 15-20 applications so only about 50 cents per application. 

 

The Gelid is my current go to paste because of the reviews I read in the past, but now I have my own evidence that lets me feel confident with that decision. 

 

One last note I think that should give a pat on the back of the Fujipoly pads is that although they performed worse than all of the other TIMs tested, they were still dealing with a 300 watt (from the wall) CPU power draw of idle versus OCCT load. The fact that this enormous load can be handled by a pad at all amazes me and I think that deserves a thumbs up even if the practicality and value isn't there. 

 

For the future I might swap a custom loop on a heavier OC like 5.1GHz and 1.5375V Vcore and see if there is a larger difference between the MX-2 and Gelid pastes. 

 

That's all for now!