Overclock.net banner
1 - 14 of 14 Posts

· Registered
Joined
·
150 Posts
Discussion Starter · #1 ·
Hi all,

So, this is my attempt at benching cooling solutions, is has been a long (several months of my spare time) and arduous process, so please be gentle.

Hardware group picture:
2520230


Here are a few points, before I dive in this monstrosity:
  • I am no GamersNexus, nor any professional reviewer, I did this as an amateur, on my own, freely, and at my own costs, I might have made mistakes.
  • I won't redo tests, nor test other coolers, this has taken me way too long to do, and I literally need a break from all of this.
  • A lot of data was scratched, for several reasons, here are a few of them:
  1. Idle data was scratched, given that Ryzen cpus turbo in and out non-stop, rendering them almost impossible to be representative across all the hardware tested, especially given that I am running on a "normal" OS.
  2. I had noise data, but, given the noise pollution (construction work near my place in the spring, and AC noise getting picked up during the summer), I refrained from including it, and even stopped recording it for a lot of tests.
  3. Humidity level was varying way too much. My region had some long periods with and without rain, which made data a bit all over the place. It might have affected results slightly, sorry for that.
  4. No, I'm not running Prime95, I am an amateur, and I don't intend on degrading my CPU running all those tests. And, lets be honest, Prime95 is an unrealistic scenario for 99.999% of people, if not more.
  5. Some tests were removed, given that consecutive run-to-run with them gave results that varied way too much to my liking, or given that they lack the duration needed for proper testing.
  6. The 2 tests that were kept are Cinebench R23 30min run, and Timespy Stability test (20 loops), representing a case of more "work-like" usage, and a case of more "gaming" usage.

The coolers that are compared in this are:
  • Coolermaster 212 Evo
  • Cryorig H7
  • Cryorig H5 Universal
  • Raijintek Tisis Core Edition
  • Noctua NH-D15S
  • 280mm AIO (NZXT X62, Asetek design)
  • 360mm AIO (Corsair H150i Pro XT, Asetek design)
  • EKWB Supremacy Evo block on a custom loop (details on that loop lower)
  • OptimusPC AMD Foundation block on a custom loop (details on that loop lower)

Setup used for all tests:
  • AMD R9 5950X (PBO and SMT Enabled, FLCK at 1900MHz)
  • Asus X570 Dark Hero (DOS (Dynamic OC switching) Enabled at 60 Amps, 4.6GHz on CCD1, 4.55GHz on CCD2, at 1.225V)
  • GSkillz Trident Z 32GB (4X8) at 3800MHz 14-14-14-26-42 1.475V
  • Lian-LI O11 Dynamic case
  • EVGA 1600T2 power supply
  • AMD RX 6900 XT with EKWB waterblock and backplate
  • Custom loop (ZMT Tubing, distro plate, two EKWB XE (60mm thick) 360mm radiators, DDC pump)
  • 6 Arctic P12 PWM PST fans (push configuration on the radiators)

Settings are as described above. Using Dynamic OC Switching allowed me to maximize CPU usage lower threaded loads (Timespy runs), whilst having a sustained, and predictable all-core load under multi-threaded loads. Under that all-core load where DOS kicked in, the power drawn by the CPU was hovering around 210-225W.

For the tests, all fans for all coolers were replaced with Arctic P12 and P14 PWM PST fans, in order to evaluate the cooling power of the heatsink solution on it's own. (YES, I did dare to replace the fans on a Noctua cooler (HERESY!), although a case could be made it might give different results with its premium stock fans.)

All tests for the AIO and custom loop were done at 100% pump speed.

Tests were conducted using PWM signal for the fans, at 25%, 50%, 75%, and 100% fan speed, as well as the included Silent, Standard, and Turbo fan curves that came on the board.

All tests were done on the same Bios and chipset driver version.

Temperature was kept between 20 and 21 degrees Celsius, in the test room, at all times (climate controlled with AC and smart thermostats).

The thermal paste used for all tests was Thermalright TFX. (14.3 W/Mk rating)
For all tests, the temperature was recorded with HWinfo64, and taken as the average temperature of the last 5 minutes of the run. (many of those tests had to be redone because I missed that mark, RIP my free time)

For AIOs and custom loop tests, given their ability to "soak" heat for quite a while before getting to an equilibrium temperature. To palliate to that, Aida64's stress test was run for 30 minutes before each AIO tests, and for 60 minutes before each custom loop test to "soak" the respective cooling solutions as much before immediately starting the tests.
For the results, temperature is in degrees Celsius, and "Crash" is used to designate any problem with the test, ranging from straight up BSOD to Massive slow down of the test due to instability by temperature or over-temperature protection kicking in.

Now, the main course, the results and analysis:

2520231


212 Evo
As it was pretty obvious to most of you, the 212 Evo is nowhere near being able to keep over 200W under control. This test was mostly to demonstrate that trying to save 20-40$ on the cooling solution, on a nearly 1000$ CPU is not the way to go. Legendary king of value air cooling has a limit.

2520232


H7
That CPU cooler we saw in every build posts a few years back. Looks great, buck, unfortunately, results are pretty similar to the 212 Evo: not enough cooling power. Note that the cooling plate is scarily small on that unit, compared to the AM4 IHS... I was very hesitant to even test it.

2520233


H5 Universal
The H5 is the first, and only, single tower cooler, in this test, being able to somewhat "tame" the 5950X whilst drawing over 200W. You need to crank the fans to ridiculous levels for that, but it can! For gaming scenarios, that cooler could actually be sufficient for some cases, and with an undervolted 5950X, I could see it working. A better pick for a single tower solution might be Noctua's famous NH-U12S, which will surely beat the H5 Universal, and be enough for the task.

2520234


Tisis Core Edition
And we have our first dual tower air cooling solution. The Raijintek Tisis Core Edition represents dual cooler solutions on the cheap. I actually got mine, brand new, for what represents 45USD (provided it came without fans though). At that performance point, any "normal" fan speed is enough to get the 5950X under control, even with a single fan! The best setup definitely is using 2 fans, as 3 fans shows little to no improvements. For the price you pay, it is the cheapest entry realistically able to deal with an overclocked 5950X.

2520235


NH-D15 (S)
The king of air coolers, the NH-D15 (S). No need for introduction here, this cooler is, arguably, the best air cooler our there. And, at sub 100 USD, it is a force to be reckoned with. It is perfectly able to maintain the 5950X under control. And, has a bonus, this is using Arctic P14 fans. Not that those are bad fans, but the stock Noctua fans are actually better, and more silent that that. So, at 100 USD, you probably get better results than what you see here! As for the Tisis, 1 fan works, but 2 fans seems to be the sweet spot, with 3 fans barely making a difference over 2 fans.

2520236


280mm AIO (NZXT X62)
Our first AIO! And, unsurprisingly, it passed with flying colors. Results are a little better than the NH-D15, so we can safely assume that a 240mm AIO would land somewhere around a NH-D15, or slightly worse than it. Going for a push-pull configuration (4 fans) improves the situation a bit, but not by a lot.

2520237


360mm AIO (Corsair 150i Pro XT)
Our second AIO fares a little bit better than the 280mm. Which is quite natural, given that it roughly has 10% more surface area to work with (360mm x 120mm vs 280mm x 140mm). Again, we see a 5950X being completely kept under control, with a little bit of improvement by going with a push-pull configuration, over a push configuration.

2520238


Prosiphon Elite
Now, the oddball of the bunch: the IceGiant Prosiphon Elite. First of all: what a freaking chungus. I mean, this thing is one heck of a unit! If you think that the NH-D15 is big, this thing will dwarf it. I was legit scared that it would be too heavy for the mobo, but it actually comes with a very sturdy mounting solution props to them for that. Now, for those not familiar with that product, it is actually a cooler using an entirely different approach to cooling as we are used to. It's neither using traditionally heatpipes, nor a pump and waterblock. Instead, this bad boy uses nifty physic to turn the difference in temperature of the liquid inside it, into a "pumping" action. Turning is into some sort of AIO, without any pump noise at all. Theoretically, the more heat output of the CPU (and thus heat you input into the Prosiphon Elite), the faster the liquid inside it will "pump", and go through its 240mm-sized, quite thick "radiator" (it's a bit more complex than that, but I am trying to simplify). Although, this thing was designed (and will work a lot better) for very high area and high heat output CPUs (cough cough... hello Threadrippers!!!), it is interesting to see how it deals with a smaller and more energy efficient 5950X. And, the results are not bad at all. It fares a little worse than a dual Tower solution, but, given that it is silent, apart from the noise of its fans, and the truly unique look and inner workings of it, it really is an interesting solution, which is actually capable of dealing with a 5950X. The downsides of it are pretty obvious though: it's massive size, and it's price, at roughly twice the price of the NH-D15. Although, for that later part, the Prosiphon Elite's price is actually reasonable, if you consider that it is a cooling solution designed for multi-thousands of dollars Threadripper CPUs.

2520239


EKWB Supremacy Evo custom loop
Now, the custom loop. Does it work? Yeah, what question is this? Is it practical from a price/performance standpoint? Not at all. But is it baller? Yep. Basically, this is just overkill, and actually needed over an hour of 100% stress test to get it to reach above 60 degrees for the Timespy test, otherwise, the sheer thermal mass of the loop could literally laugh at how little that test impacted temperatures.

2520240


Optimus PC AMD Foundation custom loop
The same can be said about this block versus the EKWB Supremacy Evo. The difference is: it's pricier, and it performs slightly better. End of story.


Conclusion:

Don't cheap out on a cooling solution for a high-end CPU. Although, you don't need to shell out stacks of cash, at the sub 100 USD mark, the NH-D15 really stands out as a very wise choice for those who want the peace of mind of air-cooling, and knowing that their CPU will be handled. If the color scheme doesn't suit you, there is a chromax black version of it, or other coolers such as the Dark Rock Pro 4 and the Assassin III will probably land rather close to it in performance. If you are looking at AIO solutions, although I didn't test it, a 240mm is probably being able to handle a 5950X, and a 280mm solution certainly will. Basically, investing part of your budget for the proper cooling of a high-TDP, high value CPU is something to take into account when purchasing components.

Anyway, I hope you people appreciated/enjoyed/found this little write-up useful. It might be a little too much for what it's worth, but it's still something I wanted to do, and it represents a lot of my free time during this Covid pandemic.

Have a nice day to you all!
 

· Registered
Joined
·
2,066 Posts
Nice work, very interesting. If I may suggest one thing, I'd rather have the temperatures as delta over ambient, to further isolate your results from ambient temperature variations.

Thanks!
 

· Registered
Joined
·
150 Posts
Discussion Starter · #3 ·
Nice work, very interesting. If I may suggest one thing, I'd rather have the temperatures as delta over ambient, to further isolate your results from ambient temperature variations.

Thanks!
I get what you mean. However, I can't really go back and record those back. However, as I mentioned, this room (as well as the rest of the house) is controlled to stay within a 1 degree range (20 to 21 degrees). So, even with those data points, results would stay the same.

The one thing that was out of my control was humidity level, though.
 

· Registered
Joined
·
2,066 Posts
I get what you mean. However, I can't really go back and record those back. However, as I mentioned, this room (as well as the rest of the house) is controlled to stay within a 1 degree range (20 to 21 degrees). So, even with those data points, results would stay the same.

The one thing that was out of my control was humidity level, though.
I think it's enough to know that it's in that degree range, thanks!

I've never consider how humidity impacts cooler performance, do you think it's measurable?
 

· Registered
Joined
·
150 Posts
Discussion Starter · #5 ·
I think it's enough to know that it's in that degree range, thanks!

I've never consider how humidity impacts cooler performance, do you think it's measurable?
I really don't know if it would affect components as much as us, but, it might be an interesting topic to approach, as technically, humidity plays in the Air's thermal density (ability to absorb heat before climbing a degree) as well.
 

· Premium Member
Joined
·
20,502 Posts
The Raijintek at half the price of the D15 does a pretty decent job cooling a 16 core. It seems from your data that it works ideally with just 2 fans.

Is it as easy to install as the D15?

Good work OP.
 

· Old to Overclock.net
Joined
·
3,518 Posts
I enjoyed it.

Good read and gives you an estimate of similar coolers. Also impressive results for the raijintek, I have that cooler go as low as 38. Now I know it is a option.
 

· Registered
Joined
·
150 Posts
Discussion Starter · #8 ·
The Raijintek at half the price of the D15 does a pretty decent job cooling a 16 core. It seems from your data that it works ideally with just 2 fans.

Is it as easy to install as the D15?

Good work OP.
It's not as easy to install as the NH-D15, but nothing too complicated either. Raijintek offers an AM4 mounting kit, but, fun fact, using the am3 mounting hardware, with the 2011 screws actually worked as a mounting option. (What I did)
 

· Registered
Joined
·
150 Posts
Discussion Starter · #9 ·
I enjoyed it.

Good read and gives you an estimate of similar coolers. Also impressive results for the raijintek, I have that cooler go as low as 38. Now I know it is a option.
That price comes with a few caveats though:
  1. You need to buy fans for it (assuming you get 2 of them, it's a 20 to 30$ investment, if you already have them, good, but it should still be accounted for in the "price" of the Tisis);
  2. You need to buy some proper thermal paste, as the one included with it is the type I would not even consider using, just from the look of it. (The one I used is on the higher end, but you can expect to shell out 5 to 10$ for this.)
Once those are accounted for, the "real" price of the Tisis Core Edition is around 65 to 80$, depending on how you kit it out. At this price point, it gets 10 to 20$ away from a NH-D15S, with its full-proof reputation. It is a really cheap solution if you already have spare thermal paste and quality fans, laying around, though.
 

· Old to Overclock.net
Joined
·
3,518 Posts
That price comes with a few caveats though:
  1. You need to buy fans for it (assuming you get 2 of them, it's a 20 to 30$ investment, if you already have them, good, but it should still be accounted for in the "price" of the Tisis);
  2. You need to buy some proper thermal paste, as the one included with it is the type I would not even consider using, just from the look of it. (The one I used is on the higher end, but you can expect to shell out 5 to 10$ for this.)
Once those are accounted for, the "real" price of the Tisis Core Edition is around 65 to 80$, depending on how you kit it out. At this price point, it gets 10 to 20$ away from a NH-D15S, with its full-proof reputation. It is a really cheap solution if you already have spare thermal paste and quality fans, laying around, though.
Yeah I have a drawer full of 120mm and 140mm fans and 2 Ziploc bags of Thermal paste. So I would still consider it.
 

· Never Finished
Joined
·
2,514 Posts
Good job bud. I have some criticisms (going into doyll mode lol):

Inlet temps weren't monitored with a thermocouple. This can give you a very accurate ambient temperature to base your delta T off of. It can potentially change the placement of a few products in your list.

O11 dynamic isn't really optimized for airflow. I can see the case performing better for radiators than with air coolers simply because there aren't fans where the glass front panel is. If you left the side panel open for the air coolers, then the air coming off of the front side fans wouldn't really reach the cooler. I think a couple of the air coolers you tested won't fit in the O11D without having the side panel off too. Testing these products on a test bench or a freer flowing case would do a better job of isolating the performance potential of the coolers from the restrictions of some internal case designs.

In my testing, coolers that let the CPU get hotter also draw more power, causing a feedback loop kinda thing. So, there is a chance that your temperatures are slightly skewed due to that. Not a biggie since that's just the natural order of things with hardware, and only crazy people like Steve normalize for power. But it's just something to think about when sampling your temperatures. Take a look at your package power when under load and you'll see a startling and frustrating correlation.

As some have pointed out, no delta temps (I'm kinda expanding on my first point). While thermostat temperatures can remain pretty consistent, that's only the temperature at the thermostat and not where the heat source is located. By the time the heat gets to the tstat, the ambient temperature at your case could be a number of degrees higher. Having the ac kick on partway through testing on some coolers can change results, especially if your room is warm from testing prior coolers. That's why its kinda important to have a thermocouple to monitor inlet temps so that you can remove this potentially disastrous variable. IR thermometers can't really be a substitute as it only reads surface temperature and not air temperature. If you're pointing at a reflective surface like a heatsink, then you won't get an accurate surface reading either. Your tolerances are likely in the 3-5C range, making the majority of your tested coolers "functionally equal" as lord Steve would say. Or maybe these smart thermostats are better than I think they are haha. I'm not exactly sure where your tstat is located.

Not saying that your results are useless since the coolers are sitting around where they should, but there are things you can do next time if you're up to it. Trust me, I know cooler testing becomes more of a pain than a part time hobby the more you go through. Hopefully this info can help any others who want to sell their soul to the cooler gods.

Oh, and thanks for keeping the fans constant and soaking the CLCs. Not many do that and is actually really nice to see. For the other people who just jump straight to the results, keep in mind that each cooler probably took around 6 hours if not more due to the number of different fanspeeds and configurations that had to be tested, not counting erroneous passes. This is time-consuming work.
 

· Registered
Joined
·
150 Posts
Discussion Starter · #12 ·
Good job bud. I have some criticisms (going into doyll mode lol):

Inlet temps weren't monitored with a thermocouple. This can give you a very accurate ambient temperature to base your delta T off of. It can potentially change the placement of a few products in your list.

O11 dynamic isn't really optimized for airflow. I can see the case performing better for radiators than with air coolers simply because there aren't fans where the glass front panel is. If you left the side panel open for the air coolers, then the air coming off of the front side fans wouldn't really reach the cooler. I think a couple of the air coolers you tested won't fit in the O11D without having the side panel off too. Testing these products on a test bench or a freer flowing case would do a better job of isolating the performance potential of the coolers from the restrictions of some internal case designs.

In my testing, coolers that let the CPU get hotter also draw more power, causing a feedback loop kinda thing. So, there is a chance that your temperatures are slightly skewed due to that. Not a biggie since that's just the natural order of things with hardware, and only crazy people like Steve normalize for power. But it's just something to think about when sampling your temperatures. Take a look at your package power when under load and you'll see a startling and frustrating correlation.

As some have pointed out, no delta temps (I'm kinda expanding on my first point). While thermostat temperatures can remain pretty consistent, that's only the temperature at the thermostat and not where the heat source is located. By the time the heat gets to the tstat, the ambient temperature at your case could be a number of degrees higher. Having the ac kick on partway through testing on some coolers can change results, especially if your room is warm from testing prior coolers. That's why its kinda important to have a thermocouple to monitor inlet temps so that you can remove this potentially disastrous variable. IR thermometers can't really be a substitute as it only reads surface temperature and not air temperature. If you're pointing at a reflective surface like a heatsink, then you won't get an accurate surface reading either. Your tolerances are likely in the 3-5C range, making the majority of your tested coolers "functionally equal" as lord Steve would say. Or maybe these smart thermostats are better than I think they are haha. I'm not exactly sure where your tstat is located.

Not saying that your results are useless since the coolers are sitting around where they should, but there are things you can do next time if you're up to it. Trust me, I know cooler testing becomes more of a pain than a part time hobby the more you go through. Hopefully this info can help any others who want to sell their soul to the cooler gods.

Oh, and thanks for keeping the fans constant and soaking the CLCs. Not many do that and is actually really nice to see. For the other people who just jump straight to the results, keep in mind that each cooler probably took around 6 hours if not more due to the number of different fanspeeds and configurations that had to be tested, not counting erroneous passes. This is time-consuming work.
First:Thank you for your lengthy reply and the appreciation of how much time it took.

Second: You have a very fair point that I should have thought about an inlet temperature probe. The O11 Dynamic was indeed used as openly as possible (top, front and side panels removes), which I totally forgot to specify. Fortunately enough, the thermostat is actually about 3 feet away from the PC, so local temperature differential might have been kept to a minimum, and the HVAC (AC) outlet is on the other side of the room. It's in no way precise enough, but, just getting that basic standardized setup was one hell of a lengthy process, which only gives me more and more admiration for tester going the extra mile (the amount of time Tech Jesus's team must spend on this is simply ridiculous). If I ever do this again (please god, no), I will try to take these into account. Having a inlet and oulet probe for both temperature and humidity, along with a very precise, and void of pollution, instrument to measure noise, would really be ideal.
 

· Registered
Joined
·
5,418 Posts
Thank you for the time put into it. It just didn't work out i had the time to test it out myself.
 

· Registered
Joined
·
248 Posts
+10 reps :)

Really nice work and very interesting to have the comparison against the custom loops. That's something that seems to be quite hard to get any reasonable data on. Thank you for taking the time to do this well and sharing your results.
 
1 - 14 of 14 Posts
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top